Benchmarking: An International Journal
Benchmarking freight transportation corridors and routes with data envelopment
analysis (DEA)
Isotilia Costa Melo, Paulo Nocera Alves Junior, Ana Elisa Perico, Maria Gabriela Serrano Guzman,
Daisy Aparecida do Nascimento Rebelatto,

Article information:
To cite this document:
Isotilia Costa Melo, Paulo Nocera Alves Junior, Ana Elisa Perico, Maria Gabriela Serrano Guzman,
Daisy Aparecida do Nascimento Rebelatto, (2018) "Benchmarking freight transportation corridors and
routes with data envelopment analysis (DEA)", Benchmarking: An International Journal, Vol. 25 Issue:
2, pp.713-742, https://doi.org/10.1108/BIJ-11-2016-0175
Permanent link to this document:
https://doi.org/10.1108/BIJ-11-2016-0175

Downloaded on: 25 April 2019, At: 11:56 (PT)
References: this document contains references to 74 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 416 times since 2018*

Users who downloaded this article also downloaded:
(2018),"Segmenting critical success factors of humanitarian supply chains using fuzzy DEMATEL",
Benchmarking: An International Journal, Vol. 25 Iss 2 pp. 400-425 <a href="https://doi.org/10.1108/
BIJ-10-2016-0154">https://doi.org/10.1108/BIJ-10-2016-0154</a>
(2018),"Efficiency ranking method using SFA and TOPSIS (ERM-ST): case of Indian banks",
Benchmarking: An International Journal, Vol. 25 Iss 2 pp. 471-488 <a href="https://doi.org/10.1108/
BIJ-08-2016-0126">https://doi.org/10.1108/BIJ-08-2016-0126</a>

Access to this document was granted through an Emerald subscription provided by emerald-
srm:478530 []

For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.

About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.

Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)

https://doi.org/10.1108/BIJ-11-2016-0175
https://doi.org/10.1108/BIJ-11-2016-0175


Benchmarking freight
transportation corridors and
routes with data envelopment

analysis (DEA)
Isotilia Costa Melo

School of Engineering of São Carlos,
Universidade de Sao Paulo, Sao Paulo, Brazil

Paulo Nocera Alves Junior
Universidade de Sao Paulo, Sao Paulo, Brazil

Ana Elisa Perico
Universidade Estadual Paulista Julio de Mesquita Filho,

Araraquara, Brazil
Maria Gabriela Serrano Guzman
Department of Production Engineering,

Universidade de Sao Paulo Escola de Engenharia de Sao Carlos,
Sao Carlos, Brazil, and

Daisy Aparecida do Nascimento Rebelatto
Universidade de Sao Paulo, Sao Paulo, Brazil

Abstract
Purpose – The purpose of this paper is to collectively measure and compare the efficiency of Brazilian
and American soybean transport corridors, from farmers to export ports, using the data envelopment
analysis (DEA).
Design/methodology/approach – This paper aims to determine routes from main producing
micro-regions to main export ports, specifically using slack-based measure and variables that represent
the three pillars of sustainability (economic, social, and environmental). The choice of variables was guided by
literature review and analyzed through the principal component analysis. After the application of the model,
the quantitative tiebreaking method of the composite index is applied.
Findings – The findings are coherent with a global report that compares soybean transportation in both
countries (Brazil and USA). Efficient routes and corridors tend to present short distance truck trips and long
distance train or barge trips. The efficiency of the inland waterway trips depends on how many barges are
used in the same expedition. Routes with more than three modes tend to be inefficient which suggest that
there is a limit for multimodality.
Originality/value – Corridor benchmarking is a rare topic in the literature and previous works normally focus
on some specific and limited corridor performance characteristics, such as cost. The main contribution of this
research is that it expands the discussion regarding corridor benchmarking and it focuses on efficiency as a whole.
The paper also proposes a method that can be applied in different logistics contexts, like expanding the study to
different countries. More specifically, this method could be used in infrastructure investments programs.
Keywords Data envelopment analysis, Freight transport, Corridor benchmarking, Route benchmarking,
Slack-based measure, Soybean
Paper type Research paper

Benchmarking: An International
Journal

Vol. 25 No. 2, 2018
pp. 713-742

© Emerald Publishing Limited
1463-5771

DOI 10.1108/BIJ-11-2016-0175

Received 21 November 2016
Revised 21 March 2017
Accepted 17 April 2017

The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1463-5771.htm

The authors would like to thank the anonymous referees for their constructive comments which
improved this paper significantly. This research is financially supported by Fundação de Amparo à
Pesquisa do Estado de São Paulo (FAPESP).

713

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



1. Introduction
The objective of this paper is to collectively measure and compare the efficiency of Brazilian
and American soybean transport corridors, from farmers to main national export ports, using
the data envelopment analysis (DEA), specifically the slack-based measure (SBM) model.

The Observatory of Economic Complexity (OEC) is a tool, developed at the MIT Media
Lab, which allows users to compose a visual analysis of countries and the products they
exchange. The OEC classifies soybean as the 44th most important product of the world
economy (in a list of 1,238 items). And shows that in 2014, the USA was responsible for
41 percent of soybean production worldwide and Brazil was responsible for 40 percent
(Simoes and Hidalgo, 2011). In other words, Brazil is one of the most relevant US competitors
in the soybean market, and presents relevant perspectives of soybean harvest growth up to
2030 (Matsuda and Goldsmith, 2009).

USA presents a consolidated infrastructure for transporting soybeans from farmers to main
American export ports (Salin, 2016). The main American producing states, Illinois, Iowa,
Minnesota, Indiana, Nebraska, Ohio, and South Dakota, were responsible for 67.47 percent of the
national soybean production in 2012. Soybean is mostly exported by ports in the Mississippi
Gulf coast (63 percent), but exportation by the western port in the Pacific North Western
Complex is also relevant (17 percent) (USDA, 2013). Unlike the USA, Brazil faces various
structural obstacles even when considering corridors from the main and most traditional
producing states: Rio Grande do Sul (RS), Paraná (PR), Mato Grosso do Sul (MS), Goiás (GO), and
Mato Grosso (MT) to the main export ports of Santos, Paranaguá, and Rio Grande (Salin, 2016).
These main producers were responsible for 75.48 percent of Brazilian soybean exports in 2015
(MDIC – Ministry of Industry, Foreign Trade and Services, 2016).

Despite being used for years as a concept, there is no precise definition for a
“transportation corridor.” The World Bank publication Best Practices in Management of
International Trade Corridors (Arnold, 2006) provides a descriptive definition that suits the
way this term is used in this paper. According to the publication, transportation corridors
have both physical and functional elements. The most relevant physical element of a
corridor is that it must include one or more routes to provide a connection between economic
centers, while others (most of them) are long and not defined by the main gateway, e.g.
a port. In most cases, corridors involve multiple modes, but they may also be unimodal.
The most relevant functional element of a corridor is that it is usually developed to support
regional economic growth.

The literature describing methods for benchmarking corridors is rare. In addition to the
World Bank report (Arnold, 2006) which addressed the benchmarking corridor, EU Super
Green Project has been the second most relevant study. This project was aimed at
advancing the green corridor concept through a benchmarking exercise involving key
performance indicators (KPIs) of 16 European corridors (Panagakos, 2016). Colombian
researchers developed a strategic freight transport model incorporating external costs, and
applied it to the seven most important national freight transport corridors (Marquez and
Cantillo, 2013). Benchmarking corridors based on the efficiency of transportation is much
rarer. Researchers developed a framework for a route selection from Thailand to Vietnam,
using the analytic hierarchy process (AHP) and DEA to evaluate multimodal risk (Kengpol
et al., 2014). Brazilian researchers have recently combined a balanced scorecard (BSC) and
DEA to benchmark efficiency of uni- and multimodal routes that originate from Mato
Grosso and go toward the export ports in Northern and Southern regions (Oliveira and
Cicolin, 2016).

In this work, the DEA-SBM model (Tone, 2001) is proposed for benchmarking corridors.
The results obtain from the DEA show a raking of the efficiency of decision-making units
(DMUs). Each corridor is a DMU, defined by variables of interest (e.g. logistic costs,
warehouse capacity, lead time, among others to be discussed under Sections 2 and 3). In this

714

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



paper, we analyze collectively American and Brazilian national corridors which originated
in from the main soybean producing states to the main export ports. The analysis of
the elements of the results (e.g. rank, slacks) enables to show which characteristics must be
enhanced for improving efficiency. This may be used as a tool for directing infrastructure
investment programs.

Corridor benchmarking is a rare topic in the literature. The aforementioned papers about
the theme are normally focused on some specific and limited performance characteristics,
such as cost. The main contribution of the present research is that it expands the discussion
about creating a method for benchmarking corridors and it is focused on efficiency as a
whole, considering the three pillars of sustainability (economic, social and environmental).
It proposes a method that can be reapplied in different logistics contexts. In particularly,
it may be used to infrastructure investment programs.

The rest of the paper is organized as follows. Section 2 presents a literature review on
route benchmarking, explains basic concepts of DEA and mathematically demonstrates the
SBM model. Section 3 outlines the methodology. The results of the application to the
aforementioned context are presented under Section 4. Section 5 analyzes/introduces
relevant discussions. Finally, Section 6 summarizes the conclusions of the paper. The data
and references used are shown in Table AI and References.

2. Literature review
Castro and Frazzon (2017) overviewed literature in order to track the most relevant
contributions on benchmarking methods and to understand the similarities among the
studies. The authors concluded that the recent increasing of production about the theme is
substantial. They identified two main clusters of co-cited articles: one study using a large
variation of benchmarking methods and another with the DEA. Castro and Frazzon (2017)
highlighted that new DEA approaches seem to address most of the criticized issues of
previous benchmarking methods.

The DEA technique is one of the most famous and widely used tools to measure
productivity and efficiency in complex problems. This technique is applied to evaluate the
performance of the examined units, called DMUs. There are several works related to DEA
application for benchmarking different route options or transportation modes. For
example, Shao and Sun (2016) analyzed performance of Chinese freight air routes;
Grigoroudis et al. (2014) used DEA to design and analyze supply chains for biomass
transportation; Sheth et al. (2013) analyzed the performance of bus routes; Tongzon (2001)
used the DEA for analyzing the efficiency of Australian ports, as well as some specific
routes. While there are many additional relevant studies, for the purpose of thematic
delimitation, only investigations directly related to the scope will be detailed in the
following subsections.

2.1 Benchmarking routes and corridors
Eight works, among reports (Arnold, 2006; Texas Transportation Institute (TTI), 2007,
2012; Salin, 2016), academic papers (Marquez and Cantillo, 2013; Kengpol et al., 2014,
Oliveira and Cicolin, 2016), and a chapter in a book (Panagakos, 2016) are known and
considered as the basis for determining the most relevant variables when studying
benchmarking routes and corridors.

The book chapter was written under the context of EU Super Green Project.
This project aimed to advance the green corridor concept through a benchmarking
exercise involving KPIs of 16 European corridors. The variables were chosen combining
the three pillars of sustainability (economic, social, and environmental). After several
attempts to develop this method, EU Super Green excluded social variables from its
benchmark method; according to the authors, it is possible that the exclusion reflects the

715

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



secondary role that stakeholders attach to social issues when it comes to freight logistics
(Panagakos, 2016).

The US Department of Transportation Maritime Administration’s publication
“A modal comparison of domestic freight transportation effects on the general public”
(TTI, 2007), elaborated by Texas Transportation Institute (TTI), provides a description of
the “standardization” method used in order to compare the three modes (rail, truck, and
inland waterways). A more recent publication (TTI, 2012) updated data considering fleet
technology improvement but did not suggest any method alteration. The report identified
the following topical study areas for its research: cargo capacity, congestion, emissions,
energy efficiency, safety impacts (divided among fatalities, injuries and hazardous
materials incidents) and infrastructure impact. According to the authors, the
selected topics were issues associated with all modes, which enable the comparison
across modes and the importance of these issues has been verified by the stakeholders.
The global annual report “The soybean transportation guide” (Salin, 2016), which is
published by the US Department of Agriculture, focuses on logistic direct costs and
accurately describes current Brazilian transport infrastructure investment programs
status. The World Bank (Arnold, 2006) report was used as a starting point but further
discarded as a reference for choosing variables because it does not consider environmental
impacts and external costs.

Marquez and Cantillo (2013) considered time and operation as internal costs, and
congestions, accidents, air pollution, and emissions as external cost. Kengpol et al. (2014)
proposed an integrated quantitative risk assessment, AHP, and DEA to evaluate the
multimodal transportation risk. Its analysis considered origin/destination, time and cost of
each route, and emissions. Oliveira and Cicolin (2016) applied the DEA, a BCC model
oriented to outputs, to evaluate the efficiency of the logistics of 17 Brazilian grain freight
routes. The BSC was used to define variables based on four dimensions: financial, customer,
internal business, and learning and growth.

Table I details contexts, tools, and variables used by each publication.

2.2 DEA: model and associated techniques
The DEA is a non-parametric mathematical programming method used to measure
relative efficiency of DMUs in a system with multiple variables, known as inputs
and outputs. In 1978, the first mathematical model of the DEA was developed. The model
was named CCR and it presented constant scale (Charnes et al., 1978). In 1984, a model
with variant scale was developed (BCC) (Banker et al., 1984). For both models,
it is mandatory to choose the direction of projection: to prioritize output maximization or
input minimization.

Real problems, such freight transportation, may require output maximization and input
minimization simultaneously. For example, in most cases, it may be equally desired to
reduce lead time and maximize cargo capacity of each trip. Furthermore, in models called
“additives,” developed in 1985 (Charnes et al., 1985), it is not necessary to choose the
direction as they will maximize output and minimize input simultaneously. Moreover,
additive models may be variant or invariant regarding scale. One of the disadvantages of
additive models is that they do not directly determine the efficiency of the DMU, and only
identify the efficient DMUs and the targets for the DMUs inefficiencies. In 2001, the SBM
was proposed due to the limitations of the additive model. The SBM is an additive variant
that maximizes output and minimizes inputs, but results in efficiency being measured on a
scale of 0-100 percent. For the SBM, the value of the objective function is the sum of the
relative distances, also called slacks. A brief mathematical demonstration of SBM model is
shown in the following paragraphs.

716

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Pu
bl
ic
at
io
n

T
yp

e
Co

nt
ex
t

T
oo
ls

V
ar
ia
bl
es

Pa
na
ga
ko
s
(2
01
6)

Ch
ap
te
r
bo
ok

E
U
Su

pe
r
G
re
en

Pr
oj
ec
t.

ke
y
pe
rf
or
m
an
ce

in
di
ca
to
rs

O
ri
gi
n-
de
st
in
at
io
n,

ty
pe

of
m
ov
ed

ca
rg
o,
m
od
al

av
ai
la
bi
lit
y,
us
ed

ro
ut
es
,t
ra
de

im
ba
la
nc
e,
am

on
g

ot
he
rs

no
t
sp
ec
ifi
ed

in
te
xt

T
ex
as

T
ra
ns
po
rt

In
st
itu

te
(T
T
I,

20
07
,2
01
2)

R
ep
or
t
fo
r
U
S
D
ep
ar
tm

en
t
of

T
ra
ns
po
rt
at
io
n
an
d
M
ar
iti
m
e

A
dm

in
is
tr
at
io
n

In
ve
st
ig
at
io
n
of

al
te
rn
at
iv
es

to
M
is
si
ss
ip
pi

W
at
er
w
ay

in
ca
se

of
dr
ou
gh

t

D
ir
ec
t
co
m
pa
ri
so
n

am
on
g
di
ff
er
en
t
m
od
es

Ca
rg
o
ca
pa
ci
ty
,c
on
ge
st
io
ns
,e
m
is
si
on
s,
en
er
ge
tic

ef
fic
ie
nc
y
of

ea
ch

m
od
e,
sa
fe
ty

im
pa
ct
s
an
d

in
fr
as
tr
uc
tu
ra
li
m
pa
ct
s

A
rn
ol
d
(2
00
6)

R
ep
or
t
fo
r
th
e
W
or
ld

B
an
k

Co
m
pr
eh
en
si
ve

re
vi
ew

of
ho
w

tr
an
sp
or
t
co
rr
id
or
s
fu
nc
tio

n
Li
te
ra
tu
re

re
vi
ew

N
ot

ap
pl
ic
ab
le

Sa
lin

(2
01
6)

A
nn

ua
lr
ep
or
t
fo
r
U
S

D
ep
ar
tm

en
t
of

A
gr
ic
ul
tu
re

V
is
ua
ls
na
ps
ho
t
of

B
ra
zi
lia
n
so
yb

ea
n

tr
an
sp
or
ta
tio

n.
D
at
a:
th
e
co
st
of

sh
ip
pi
ng

so
yb

ea
ns

vi
a

hi
gh

w
ay
s
an
d
oc
ea
n

In
fo
rm

at
io
n
ab
ou
t
so
yb

ea
n
pr
od
uc
tio

n,
ex
po
rt
s,
ra
ilw

ay
s,
po
rt
s,
an
d

in
fr
as
tr
uc
tu
ra
ld

ev
el
op
m
en
ts

Su
m
m
ar
y
of

up
da
te
d
da
ta

N
ot

ap
pl
ic
ab
le

M
ar
qu

ez
an
d

Ca
nt
ill
o
(2
01
3)

A
rt
ic
le

T
o
de
te
rm

in
e
th
e
in
te
rn
al

an
d
ex
te
rn
al

co
st
s
of

Co
lo
m
bi
an

tr
an
sp
or
t
ro
ut
es
,

co
ns
id
er
in
g
th
re
e
m
od
es

(r
oa
d,

ra
il
an
d

in
la
nd

w
at
er
w
ay
)

M
od
el
de
ve
lo
pe
d
by

Co
lo
m
bi
an

in
st
itu

tio
n

T
im

e
an
d
op
er
at
io
ns

(in
te
rn
al
co
st
s)
,c
on
ge
st
io
n,

ac
ci
de
nt
s,
ai
r
po
llu

tio
n
an
d
ac
ci
de
nt
s

(e
xt
er
na
lc
os
ts
)

K
en
gp

ol
et
al
.

(2
01
6)

A
rt
ic
le

T
o
de
ve
lo
p
a
to
ol

th
at

al
lo
w
s
th
e

se
le
ct
io
n
of

th
e
ro
ut
e
w
ith

le
ss

ri
sk

of
lo
ad

lo
ss

be
tw

ee
n
T
ha
ila
nd

an
d
Ch

in
a

T
oo
ld

ev
el
op
ed

w
ith

th
e

ap
pl
ic
at
io
n
of

D
E
A

an
d
A
H
P

T
im

e,
co
st

of
fr
ei
gh

t,
em

is
si
on
s

O
liv

ei
ra

an
d

Ci
co
lin

(2
01
6)

A
rt
ic
le

T
o
co
m
pa
re

ef
fic
ie
nc
y
of

co
rn

flo
w

ro
ut
es

fr
om

M
at
o
G
ro
ss
o
to

th
e
m
ai
n

B
ra
zi
lia
n
po
rt
s

Ch
oi
ce

of
va
ri
ab
le
s
w
ith

ba
la
nc
ed

sc
or
ec
ar
d

(B
SC

),
ro
ut
e
be
nc
hm

ar
k

w
ith

D
E
A
-B
CC

Lo
gi
st
ic
s
co
st
s,
tim

e,
em

is
si
on
s,
tr
an
sp
or
ta
tio

n
m
at
ri
x,

av
ai
la
bi
lit
y
of

w
ar
eh
ou
se
s,
co
st

of
co
rn

pr
od
uc
tio

n,
lin

e
in

po
rt
/s
hi
p,

ro
ut
e
ex
te
ns
io
n,

tr
an
sp
or
t
co
st
s,
tr
an
sp
or
t
sp
ee
d,

fu
el

co
ns
um

pt
io
n
an
d
fle
et

ag
e

Table I.
Summary of

literature review

717

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



In this paper, routes were defined as DMUs. Dealing with n DMUs with inputs x and outputs
y, it is assumed that the data are positive, i.e. xW0 and yW0. Consider an expression to
describe a certain DMU (x0·y0) as:

xi0 ¼ xiklkþs�i

yr0 ¼ yrklk�sþr

with λk⩾ 0, s�i X0, and sþr X0. The λk indicates the contribution of the kth DMU; s�i and sþr
indicate, respectively, the excesses of ith input and rth output. Using s�i and sþr , the index ρ
for efficiency can be defined as:

r ¼ 1� 1=m
� �Pm

i¼1 s
�
i =xi0

1þ 1=s
� �Ps

r¼1 sþr =yr0
:

In an effort to estimate the efficiency for (x0·y0), the creator of the SBM model (Tone, 2001)
established the fractional programming in λk, s�i , and sþr , with the next step being
linearization. The author multiplies the numerator and the denominator of the function by a
positive number, t, in order to achieve linearization. This process sets the denominator to be
equal to 1 and moves it to the constraints. The goal is to minimize the numerator. The
problem remains non-linear, once the variables are multiplied by t. However, it is possible to
make it linear by transforming each variable multiplied by t into a single new variable, i.e.
S�
i ¼ ts�i , S

þ
r ¼ tsþr , and Λk¼ tλk.

In order to insert non-discretionary variables (also known as non-controllable or
environmental factors) into SBM models, such as the extension route variable used in this
study, consider the following constraints of the extended additive model (Charnes et al.,
1987), as recommended by previous works (Saen, 2005):

S�
i pbixi0

Sþ
r pgryr0

The βi and gr are the prescribed parameters between 0 and 1 that indicate the degree of
discretion of each input or output. If the variable is completely non-controllable, the
parameter is equal to 0. If the variable is completely discretionary, the parameter is equal to
1. If the variable is partially under control, the parameter has a value between 0 and 1. If the
variables are freely discretionary, parameters → + ∞, the constraint is removed and it
becomes a standard SBM model.

The SBM formulation as a linear programming model with non-discretionary variables is
as follows:

Minimize t ¼ t� 1=m
� �Xm

i¼1

S�
i =xi0 (1)

subject to:

1=s
� �Xs

r¼1

Sþ
r =yr0þ t ¼ 1 (2)

718

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Xm

i¼1

LkxikþS�
i �txi0 ¼ 0 k ¼ 1; 2; . . .; z (3)

Xs

r¼1

Lkyrk�Sþ
r �tyr0 ¼ 0 k ¼ 1; 2; . . .; z (4)

S�
i pbixi0 i ¼ 1; 2; . . .; p (5)

Sþ
r pgryr0 r ¼ 1; 2; . . .; q (6)

LkX0; S�
i X0; Sþ

r X0 and t40 (7)

where τ is the efficiency; S�
i the slack of the ith input; Sþ

r the slack of the rth output; Λk the
contribution of the kth DMU; t the model linearization factor; xi0 the ith input of DMU under
analysis; yr0 the rth output of the DMU under analysis; xik the ith input of the kth DMU; yrk
the rth output of the kth DMU; m the number of inputs; s the number of outputs; p
the number of non-discretionary inputs; q the number of non-discretionary outputs; and z
the number of DMUs.

It is possible to impose some restrictions forΛk, such as
Pz

k¼1 Lk ¼ 1 (variable returns to
scale). The optimum solution (ρn, tn, Lk

n, Si
�n, Sr

þn) is given by:

rn ¼ tn; lnk ¼ Ln

k=t
n; s�n

i ¼ S�n

i =tn; sþn

r ¼ Sþn

r =tn

Based on this solution, it is defined that a DMU (x0·y0), defined as a route in this work, is
efficient in an SBM model when ρn¼ 1.

A few of the issues among the problems related to variables are: undesirable outputs
and excessive number of variables to relatively few DMUs. The maximization of some
outputs, such as emissions, may not be desired, as they have negative impacts on the
environment and/or on society (Camioto et al., 2014). These outputs are called undesirable.
It is recommended that they be inserted as input in order for them to be minimized
(Cook et al., 2014).

The goal of the Principal component analysis (PCA)-DEA model is to improve the
discriminatory power of the DEA, which often fails when there is an excessive number of
inputs and outputs in relation to DMUs (Adler and Golany, 2007). The PCA explains the
variance structure of a matrix data through a linear combination of variables. Consequently,
it may be possible to reduce data to a few principal components, which generally describes
80-90 percent of the variance in data. If most of the population variance can be attributed to
the first few components, then they can replace the original variables with minimum loss of
information (Hair et al., 1995). The PCA used here is based on correlation rather than
covariance, as the variables used in the DEA are often quantified in different units of
measure (Adler and Golany, 2007). The PCA-DEA is widely used in international research
and, in comparison with variable reduction based on partial covariance (VR), presents more
robust and accurate results (Adler and Yazhemsky, 2010).

A common problem found in DEA rankings is the number of ties identified as efficient
DMUs. This paper applied the composite index tiebreaking method (Leta et al., 2005). This
method considers the average between standard and inverted efficiencies, and later
standardizes them by the maximum index.

In this paper, standard efficiency Estandard
k is the efficiency τn calculated by the SBM

model. The inverted efficiency E inverted
k is calculated by handling inputs as outputs, and

719

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



outputs as inputs (Mariano and Rebelatto, 2014; Leta et al., 2005). The formulation of the
composite index is as follows:

Ecomposite
k ¼

Estandard
k þ 1�E inverted

k

� �h i
=2

max Estandard
k þ 1�E inverted

k

� �h i
=2

n o k ¼ 1; 2; . . .; z (9)

where Ecomposite
k is the composite index of the kth DMU; Estandard

k the standard efficiency
of the kth DMU; E inverted

k the inverted efficiency of the kth DMU; and z the number of DMUs.

3. Methods
3.1 Definitions of routes and corridors
The corridors were defined considering the information from the harvest of 2014 (a year not
affected by extraordinary climatic events). The routes were established from the top three
producing micro-regions to the top export ports.

In Brazil, official sources (ANTAQ, 2011; EMATER – State Company of Technical
Assistance and Rural Extension, 2014; IMBEES – Institute Mauro Borges of Statics and
Socioeconomic Studies, 2015; IMEA – Institute of State of Mato Grosso of Agricultural
Economics, 2015; SEAB – Secretariat of Agriculture and Supply, 2015; SIGA –
Agribusiness Geographic Information System, 2015; IBGE – Brazilian Institute of
Geography and Statistics, 2016a), companies (AHRANA – Administration of Paraná
Waterways, 2005, 2012, 2013; Alianca – Trevisa Investments, 2016; ALL, 2016;
Ferroeste – Secretariat of Infrastructure and Logistics, 2016), and previous works
(Vieira, 2002; Ojima, 2004; Rocha and Parré, 2009; Oliveira and Cicolin, 2016) were
consulted. The reports of the Consulting Company Macrologística (commissioned by the
National Federation of Industry) were also consulted (Macrologística, 2011, 2013).
In all, 19 corridors and 72 routes were identified. In the USA, official sources (Casavant
et al., 2011; Salin, 2016; NASS – National Agricultural Statistics Services (NASS), 2016;
NOAA, 2016; US Corps of Engineers, US Army, 2016; Waterways Council, 2016) were
consulted. Ten corridors and 30 routes were identified.

All the routes assigned a code, according to the existing modes of transportation and the
country. The codes of all the American routes begin with an “E”. If there is road
transportation, the code receives the letter “R”. If there is barge transportation, the code
receives the letter “H”. If there is a rail transportation, the code receives the letter “F”.
Table II summarizes all routes and corridors.

3.2 Proposal of variables
Based on the information provided, the initial proposal of variables was: fuel consumption and
planted area as inputs; transported harvest and in-farm static storage capacity as outputs.
There are also undesirable outputs, such as fatalities, off-farm static storage capacity,
emissions, and disposal factor. As they were undesirable, they were inserted as inputs.
The reasoning as to why the variable route extension was considered non-discretionary will
be discussed in the next few paragraphs. Table III summarizes how the variables were chosen,
their sources, their rationality, and previous works that used the same variables.

3.2.1 Fuel consumption. The input “fuel consumption” was designed to represent the
cost of transport. Different countries have different currencies, economic structures, and
taxes. This context hinders the direct comparison of costs, e.g. even by converting the
costs to the same currency, there may be discrepancies such as different incurred taxes,
labor costs, etc. Diesel is an input consumed in transport by all modes in both countries.

720

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Moreover, it is derived from oil, the price of which is governed by international quotations.
Thus, diesel consumption facilitates the representation and comparison of costs between
different countries.

The dimensions of the units used to transport freight vary widely within each of the
three modes (rail, truck, and inland waterway). They also vary according to the
availability of technology in each country. In order to build a meaningful cross-modal
comparison, “standard” dimensions of the units used by each mode were defined. In this
manner, all three modes were evaluated on the same scale, allowing the use of the DEA for
benchmarking routes.

Corridors Routes
Code Origin Destination Mode Code

C1 RS Rio Grande Road R1, R2, R3
Rail F1
Road and waterway RH1, RH2, RH3, RH4, RH5, RH6, RH7, RH8,

RH9
Road and rail RF2
Rail and waterway FH1
Road, rail and waterway RFH1

C2 PR Paranaguá Road R4, R6, R8
Rail F2, F3, F4
Road and rail RF1

C3 PR Santos Road R5, R7, R9
C4 MS Paranaguá Road R10, R12, R14
C5 MS Santos Road R11, R13, R15

Rail F5
Road and rail RF3, RF4

C6 GO Paranaguá Road R16, R18
Road and rail RF5, RF6

C7 GO Santos Road R17, R19, R20
Road, waterway and
road

RHR1, RHR2, RHR3, RHR4, RHR5, RHR6

Road, waterway and rail RHF1, RHF2, RHF3
C8 MT Paranaguá Road R21, R23, R25

Road and rail RF7, RF8, RF9
C9 MT Santos Road R22, R24, R26

Road and rail RF10, RF11, RF12
Road, waterway and rail RHR7, RHR8, RHR9, RHR10, RHR11, RHR12
Road, waterway and rail RHF4, RHF5, RHF6

EC1 MO Gulf of
Mississippi

Barge EH1

Road and waterway ERH13, ERH14
EC2 MO NPW Road and rail ERF13, ERF14, ERF15
EC3 IL Gulf of

Mississippi
Road and waterway ERH1, ERH2, ERH3

EC4 IL NPW Road and rail ERF1, ERF2, ERF3
EC5 IA Gulf of

Mississippi
Road and waterway ERH4, ERH5, ERH6

EC6 IA NPW Road and rail ERF4, ERF5, ERF6
EC7 IN Gulf of

Mississippi
Road and waterway ERH7, ERH8, ERH9

EC8 IN NPW Road and rail ERF7, ERF8, ERF9
EC9 MN Gulf of

Mississippi
Road and waterway ERH10, ERH11, ERH12

EC10 MN NPW Road and rail ERF10, ERF11, ERF12

Table II.
List of considered

routes and
respective corridors

721

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



V
ar
ia
bl
e

Pi
lla
r
of

su
st
ai
na
bi
lit
y

T
yp

e
So
ur
ce
s

R
at
io
na
l

Pr
ev
io
us

w
or
ks

Fu
el
co
ns
um

pt
io
n

E
co
no
m
ic

In
pu

t
Ca

lc
ul
at
ed

ac
co
rd
in
g
to

T
T
I
(2
00
7)
,T

T
I

(2
01
2)

It
re
pr
es
en
ts

lo
gi
st
ic
s

co
st
s
in

bo
th

co
un

tr
ie
s.

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

Pl
an
te
d
ar
ea

E
co
no
m
ic

In
pu

t
Co

lle
ct
ed

fr
om

N
A
SS

(2
01
6)
,C

O
N
A
B
–

N
at
io
na
lS

up
pl
y
Co

m
pa
ny

(2
01
4)
.

It
di
re
ct
ly

af
fe
ct
s

tr
an
sp
or
te
d
ha
rv
es
t

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

T
ra
ns
po
rt
ed

ha
rv
es
t
E
co
no
m
ic

O
ut
pu

t
Co

lle
ct
ed

fr
om

N
A
SS

(2
01
6)
,C

O
N
A
B
–

N
at
io
na
lS

up
pl
y
Co

m
pa
ny

(2
01
4)

It
w
as

as
su
m
ed

al
l

ha
rv
es
t
is
tr
an
sp
or
te
d

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

In
-fa

rm
st
at
ic

st
or
ag
e
ca
pa
ci
ty

E
co
no
m
ic

O
ut
pu

t
Co

lle
ct
ed

fr
om

N
A
SS

(2
01
6)
,I
B
G
E
–
B
ra
zi
lia
n

In
st
itu

te
of

G
eo
gr
ap
hy

an
d
St
at
is
tic
s
(2
01
6b
)
B
ra
zi
lp

re
se
nt
s

w
ar
eh
ou
se

de
fic
it

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

Fa
ta
lit
ie
s

So
ci
al

U
nd

es
ir
ab
le
ou
tp
ut

E
st
im

at
ed

ba
se
d
on

N
at
io
na
lH

ig
hw

ay
T
ra
ff
ic
Sa
fe
ty

A
dm

in
is
tr
at
io
n
(2
01
6)
,F

ed
er
al

R
ai
lr
oa
d
A
dm

in
is
tr
at
io
n
O
ff
ic
e
of

Sa
fe
ty

A
na
ly
si
s
(2
01
6)
,B

ur
ea
u
of

T
ra
ns
po
rt
at
io
n

St
at
is
tic
s
(2
01
6)
,D

ep
ar
ta
m
en
to

N
ac
io
na
ld

e
In
fr
ae
st
ru
tu
ra

de
T
ra
ns
po
rt
es

(D
N
IT
,2
01
0)
,

IM
T
T
–
In
st
itu

te
of

M
ob
ili
ty

an
d
La

nd
T
ra
ns
po
rt
(2
01
1)
,a
nd

Fe
rr
ei
ra

(2
01
0)

R
oa
d
tr
an
sp
or
ta
tio

n
pr
es
en
ts

a
hi
gh

er
le
ve
l

of
fa
ta
lit
ie
s
th
an

ot
he
r

m
od
es

T
T
I(
20
07
),
T
T
I(
20
12
),
M
ar
qu

ez
an
d
Ca

nt
ill
o
(2
01
3)

D
is
po
sa
lf
ac
to
r

E
nv

ir
on
m
en
ta
l
U
nd

es
ir
ab
le
ou
tp
ut

Ca
lc
ul
at
ed

ba
se
d
on

si
ze

an
d
ag
e
of

fle
et

ba
se
d
on

A
B
IP
E
CA

S
–
B
ra
zi
lia
n
A
ss
oc
ia
tio

n
of

A
ut
om

ot
iv
e
Pa

rt
In
du

st
ri
es

(2
01
1)
,A

lia
nc
a

–
T
re
vi
sa

In
ve
st
m
en
ts

(2
01
6)
,N

ev
es

(2
01
2)
,

B
ur
ea
u
of

T
ra
ns
po
rt
at
io
n
St
at
is
tic
s
(2
01
6)
,

M
ur
ra
y
(2
01
6)
,S

T
A
T
IS
T
A

(2
01
2)

It
re
pr
es
en
ts

th
e

di
sp
os
al

of
pr
od
uc
tiv

e
as
se
ts

af
te
r
th
e
en
d
of

th
e
ec
on
om

ic
cy
cl
e

N
ot

us
ed

by
pr
ev
io
us

w
or
ks
,b
ut

its
m
ai
n
co
m
po
ne
nt

is
fle
et

ag
e,

co
ns
id
er
ed

by
O
liv

ei
ra

an
d

Ci
co
lin

(2
01
6)

E
m
is
si
on
s

E
nv

ir
on
m
en
ta
l
U
nd

es
ir
ab
le
ou
tp
ut

Ca
lc
ul
at
ed

ac
co
rd
in
g
to

G
re
en
ho
us
e
G
as

Pr
ot
oc
ol

–
W
or
ld

R
es
ou
rc
es

In
st
itu

te
(2
01
6)

R
ed
uc
in
g
em

is
si
on
s
is
a

co
nc
er
n
fo
r
m
os
t

co
un

tr
ie
s
(C
am

io
to

et
al
.,
20
14
)

Pa
na
ga
ko
s
(2
01
6)
,T

T
I
(2
00
7)
,

T
T
I
(2
01
2)
,O

liv
ei
ra

an
d
Ci
co
lin

(2
01
6)
,M

ar
qu

ez
an
d
Ca

nt
ill
o

(2
01
3)
,K

en
gp

ol
et
al
.(
20
14
)

O
ff
-fa

rm
st
at
ic

st
or
ag
e
ca
pa
ci
ty

E
co
no
m
ic

U
nd

es
ir
ab
le
ou
tp
ut

Co
lle
ct
ed

fr
om

N
A
SS

(2
01
6)
,I
B
G
E
–
B
ra
zi
lia
n

In
st
itu

te
of

G
eo
gr
ap
hy

an
d
St
at
is
tic
s
(2
01
6b
)
B
ra
zi
lp

re
se
nt
s

w
ar
eh
ou
se

de
fic
it

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

R
ou
te

ex
te
ns
io
n

N
ot

ap
pl
ic
ab
le

N
on
-d
is
cr
et
io
na
ry

Co
lle
ct
ed

fr
om

G
oo
gl
e
M
ap
s
(2
01
6)

It
di
re
ct
ly

af
fe
ct
s
fu
el

co
ns
um

pt
io
n,

bu
t
it
is

no
t
un

de
r
to
ta
lc
on
tr
ol

of
de
ci
si
on
-m

ak
er
s

O
liv

ei
ra

an
d
Ci
co
lin

(2
01
6)

Table III.
Summary of
variable choice,
source and rational

722

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



For the calculation of fuel consumption, Baumel (2011) and Baumel et al. (2015) compared
several methods, including the proposed by TTI (2007), and proved that the best way to
estimate fuel consumption for bulk freight transportation is specifically collecting data from
each route. However, due to the purpose of this paper, it was not physically possible to
locally collect data.

Fuel consumption was also calculated by using the TTI method once it was accepted
by the Soy Transportation Coalition (2016). For calculation according to the TTI method,
it was important to transform cargo unit capacity of the vehicles in each country, as can be
seen in Table IV.

For the USA, the energy efficiency of each mode was calculated as presented in the last
report (TTI, 2012). For Brazil, the energy efficiency was calculated using data from the
current condition of the Brazilian freight fleet.

In Brazil, the typical bulk commodity truck’s body type, axle configuration, fuel, gross,
tare, and cargo weight used in this paper were confirmed by Conselho Nacional de Trânsito.
The truck’s body type used in this paper was a CVC (Combinação de Veículo de Carga,
Combination of Cargo Vehicles in Portuguese), class 3D4, code 88, also known as Romeo and
Juliet (a combination of a tractor and two trailers), with a Gross Vehicle Weight Rating of
57 tons, which includes 43 tons of cargo weight. The typical axle configuration is legally
determined as a tractor with two/three axles, in which two are for steering, commonly
named “6× 4”. The typical trailer configuration has two tandem axles. The CVC has total
seven axles (CONTRAN – National Traffic Counsil, 1999).

In Brazil, the specification regarding the typical railcar for carrying bulk commodities
was confirmed by the main rail company (ALL – Latin America Logistics, 2016) and
national specialized publications (Neves, 2012). For information regarding the barges,
two inland waterways’ complexes were analyzed: the Paraná Waterway and Lake of
Ducks. The Paraná Waterway is limited by the structural project to only two
combinations of barges: type Paraná (three barges and a tow) and type Tietê (two barges
and a tow). It was assumed that all companies operate with the combination type Paraná.
An arithmetic average of barge cargo capacity of all operating companies was calculated
(AHRANA – Administration of Paraná Waterways, 2012). This was assumed as the
“standard” barge and the “standard” combination. The Lake of Ducks is a fluvial lake
without known structural restrictions. The company Aliança is the oldest and main
operator (Alianca – Trevisa Investments, 2016). Its fleet was analyzed and it was similar to
the “standard” barge from the Paraná Waterway; therefore the same characteristics were
assumed for barges in both inland waterways.

In resume, TTI (TTI, 2007) developed a method for comparing different transport modes
applied to a study in the USA. The differences between American and Brazilian freight transport
characteristics are principally the cargo capacity, fleet combination, and engine power. These
characteristics directly affect fuel consumption, fuel emissions, and volume of cargo carried trip.

Brazil USA

Modal freight unit
Standard cargo

unit capacity (ton)
Standard fleet
combination

Standard cargo
unit capacity (ton)

Standard fleet
combination

Highway – truck
trailer

43.0 1 tractor (6× 4):2 trailers
(total: seven axes)

25.0 1 tractor (6× 2) :
trailer (total: six axes)

Barge – dry bulk 1287.0 3 barges : 1 barge-tow 1620.0 15 barges : 1
barge-tow

Rail – bulk car 80 80 cars : 4 locomotives 110 108 cars : 3
locomotives

Table IV.
Standard modal

freight unit capacity
and standard fleet

combination
(per country)

723

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



3.2.2 Warehouses. Oliveira and Cicolin (2016) considered the availability of warehouses;
nevertheless, the different warehouse sizes and types with various static storage capacities
were not considered. This paper considers the sum of static storage capacity by state,
divided between in-farm and off-farm warehouses. “In-farm static storage capacity”
considers the warehouses that are located inside farms. This capacity normally presents less
expensive storage costs, since it falls under the property of the farmer, and the space belong
to the farmer. The producers may stock soybean for several months waiting for the best
price of sale. The storage may reduce freight transportation flow from farms to export ports
during the harvest season.

When considering in-farm static storage capacity as a percentage of the national crop
production of each country, Brazil presents, in average, the lowest capacity (11.3 percent).
The American average is 65 percent, the European 50 percent, the Argentinian 40 percent, and
the Canadian exceeds 80 percent (CONAB – National Supply Company, 2005). Banco Nacional
do Desenvolvimento Social (BNDES), or The Brazilian National Bank for Development,
allegedly distributed five billions reais (Brazilian currency) between 2013 and 2014 in order to
expand national storage capacity, through five official programs; it is expected to distribute
more than 20 billions reais in the next few years. The effectiveness of these programs is not
yet clear; in fact, one of the most important contributions of the ongoing research on the topic
would be to define a way to measure their effectiveness, according to the BNDES analysts
(Maia et al., 2013). Due to the reasons discussed above, this paper considered the maximization
of in-farm static storage capacity desirable and classified it as an output, while “off-farm static
storage capacity” is considered an inputs as it is to be minimized. It is important to keep both
storage capacities because they involve different players. In-farm capacity is developed by
farmers, while off-farm capacity is developed by trade companies and co-ops and they may be
the focus of different funding programs.

3.2.3 Disposal factor. The “disposal factor” represents the disposal of productive assets
after the end of the majority of economic cycles. Fleet age, considered by Oliveira and Cicolin
(2016), is a strong element of the disposal factor. However, disposal factor is more embracing
because it considers the regional fleet in comparison to the national fleet so that it is possible
to compare different micro-regions in different countries.

The calculation of the disposal factor is explained in the following equation.
For multimodal routes, the total disposal factor is the sum of each mode factors:

FD ¼ CEð Þ
IFnð Þ:

TFrð Þ
TFnð Þ (10)

where FD is the disposal factor; CE the duration of the economic cycle in years, adopting
100 years; IFn the average age of the national fleet; TFr the fleet size required for the
transportation of regional crop; and TFn the fleet size required for the transportation of the
national harvest.

3.2.4 Brief considerations about other variables. The input “planted area” directly affects
the output “transported harvest.” The same planted area does not mean the same harvest,
because micro-regions may have different productivities. Oliveira and Cicolin (2016)
considered productivity, i.e. harvest divided by planted area. As harvest is already an
output in the present paper, it was preferred not to use productivity as an input, because it is
the quotient of harvest by planted area.

As it is presented under Section 1, soybeans represent 10 percent of Brazilian
exportation. For this model, it was assumed that all collected becomes “transported harvest”
sent to ports. It is an overestimation because some part of the harvest is processed near the
farms, some part is transported to other harbors (not considered by this model), and some is
lost during transportation. This overestimated assumption is valid for long-term decisions

724

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



because Brazil and the USA present significant perspectives of soybean harvest growth up
to 2030 (Matsuda and Goldsmith, 2009).

Social variables used in the comparison of the two countries were problematic since they
have different laws regarding work and pollution, and different concepts of incidents and
accidents at work (Salin, 2016). “Fatality” is a common term in both countries, i.e. death at
work, except suicide. The average of fatalities per mode per country multiplied by the route
extension of each mode is an initial attempt for the inclusion of social dimension. Dubey et al.
(2017) executed a systematic literature review about sustainable or green supply chain
management and identified its impact on the health of human beings as a possibility for
further study. Panagakos (2016) used the sum of fatalities and serious injuries per year per
million ton-km; however, the author did not consider fatalities per mode. The differences in
fatalities per mode must be considered as road fatalities are significantly higher than in any
other mode; e.g. in 2009, there were 3,236 fatalities on federal Brazilian roads, considering
exclusively routes from farms to ports (DNIT, 2010), 22 fatalities on railways
(IMTT – Institute of Mobility and Land Transport, 2011) and no registered fatality in
inland waterways (Ferreira, 2010). The number of fatalities on roads was probably
underestimated because it did not consider fatalities on state and district roads.

The variable “route extension” directly affects fuel consumption and, indirectly
emissions and fatalities. It is desirable to minimize the route extension variable, but it is
hardly possible. In both countries, the main farms are in centralized locations, far from the
ports, due to agricultural conditions, and historical, cultural and economic processes.
In most cases, government investments are assumedly not able to change their physical
structure. Therefore, route extension is assumed to be non-discretionary (non-controllable).

All variables used are listed under Table AI.

4. Results
4.1 Data analysis
The PCA was performed with SPSS software, considering the initial proposal of the
variables. Table V shows the results from the PCA (commonalities).

The communalities indicate the proportion of the variance explained by principal
components. There are as many communalities as there are variables in the model and their
values can vary between 0 and 1. If the value is 0, the communality factors do not explain
any variance of the variable. If the value is 1, they explain themselves entirely. In addition,
any value below 0.5 suggests the exclusion of the variable. Therefore, at this point, none of
the proposed variables should be deleted.

Table VI presents the eigenvalues, the percentage of variance that the factors are able to
explain and the accumulated percentage of this variance. In the last three columns, the
values of the held factors in the analysis (after extraction) are repeated and the excluded

Variables Initial Extraction

Fuel consumption (input) 1.000 0.924
Planted area (input) 1.000 0.907
Route extension (input) 1.000 0.623
Emissions (output – undesirable) 1.000 0.899
Disposal factor (output – undesirable) 1.000 0.637
Off-farm static storage capacity (output – undesirable) 1.000 0.718
Fatalities (output – undesirable) 1.000 0.583
Transported harvest (output) 1.000 0.818
In-farm static storage capacity (output) 1.000 0.868

Table V.
Communalities of

proposed variables

725

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



values are omitted. Table VI contains only three values with eigenvalues greater than 1,
which explains 77.53 percent of the total variability. Given the aforementioned results, none
of the proposed variables were excluded.

4.2 DEA-SBM model results
The model initially resulted in 58 ties of efficient DMUs. In all, 25 DMUs were efficient by
SBM and were also on the inverted frontier. After the application of the standardized
composite index, there were no additional efficient ties. Table VII presents the route ranking
results. The most efficient route (EH1) is an American unimodal route, exclusively by barge,
that links New Madrid, Missouri State, to Gulf of Mississippi. The most efficient Brazilian
routes were in the Southern region.

The composite index considers standard and inverted efficiency (Equation (9)). Among
the 58 tied DMUs, 25 also presented a value of 1.000 for inverted efficiency. That means they
are simultaneously benchmarks for good and bad practices. Some variables’ values are very
good, but the other values are extremely undesired. They could be named “weakly” efficient
DMUs. Table VIII presents simultaneously tied DMUs (weakly and strongly efficient).

Table IX presents the benchmark DMUs for each inefficient DMU, i.e. their λkW0. In all,
27 DMUs are benchmarks for the 44 inefficient DMUs. Among the benchmarks, 19 are
strongly efficient. The most influential benchmarks and the number of influenced DMUs
are: ERH12 (18 DMUs may consider it as efficiency benchmark), EH1 (17), ERH9 (15), ERH4
(11), ERF14 (11), F3 (11), RF4 (10) and RF2 (9).

It is possible to determine the impact of the variables in each DMU performance by
investigating their slacks. When analyzing inputs, the number of DMUs with slack that is
equal to 0 (i.e. it is not necessary to improve those variables to achieve benchmarks) are
58 (regarding to fuel consumption), 79 (regarding to planted area), 58 (emissions),
58 (disposal factor), 62 ( fatalities), and 86 (off-farm storage capacity). Table X shows the
percentage of necessary improvement for each variable to achieve benchmarks.

All kinds of routes (except Brazilian routes with exclusive rail transportation) present at
least one DMU with slacks due to fuel consumption, which means it may be reduced. Table X
presents the suggested percentages of alteration suggested in order to achieve benchmarks
Neither American routes nor Brazilian routes with exclusively rail transportation had slacks
for a planted area. The results suggest it is possible to reduce a planted area without reducing
production. No route exclusively by waterway, rail, or even the combination of both modes of
transportation had slacks for emissions or disposal factor. The other types of routes may
reduce those variables in order to increase efficiency. No American routes with transportation
by waterway and no Brazilian routes without road transportation present slacks for fatalities.
Due to ethical considerations, the reduction of this variable must be a priority. No route
exclusively by waterway, rail, or the combination of both modes presents slacks for off-farm
storage capacity. For other routes, the results suggest the variable may be reduced; however,

Component Eigenvalue % of variance Cumulative % Total % of variance Cumulative %

1 3.600 40.003 44.003 3.600 40.003 44.003
2 1.829 20.325 60.327 1.829 20.325 60.327
3 1.548 17.203 77.530 1.548 17.203 77.530
4 0.943 10.476 88.006
5 0.654 7.263 95.269
6 0.245 2.722 97.991
7 0.108 1.200 99.191
8 0.058 0.643 99.834
9 0.015 0.166 100.000

Table VI.
Total variance
explained of
components and
eigenvalues

726

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Co
de

R
an
k

E
ff
ic
ie
nc
y

Co
de

R
an
k

E
ff
ic
ie
nc
y

Co
de

R
an
k

E
ff
ic
ie
nc
y

Co
de

R
an
k

E
ff
ic
ie
nc
y

E
H
1

1
1.
00
00

R
H
R
7

26
0.
89
22

R
16

51
0.
50
09

R
24

76
0.
19
01

FH
1

2
0.
99
92

E
R
F1

4
27

0.
86
88

E
R
F9

52
0.
50
09

R
23

77
0.
17
36

F5
3

0.
99
85

E
R
H
4

28
0.
84
47

E
R
F5

53
0.
50
09

R
4

78
0.
16
06

F1
4

0.
99
82

E
R
H
6

29
0.
83
26

E
R
H
2

54
0.
50
09

R
F1

79
0.
09
79

F2
5

0.
99
78

R
F2

30
0.
81
68

R
17

55
0.
50
09

R
5

80
0.
08
35

F4
6

0.
99
78

E
R
H
5

31
0.
75
13

R
11

56
0.
50
09

R
8

81
0.
07
17

F3
7

0.
99
78

E
R
F7

32
0.
74
32

E
R
F6

57
0.
50
09

R
H
3

82
0.
07
15

E
R
H
12

8
0.
99
71

E
R
F8

33
0.
73
18

E
R
F1

5
58

0.
50
09

R
7

83
0.
06
90

R
H
R
12

9
0.
99
39

R
H
F1

34
0.
71
71

R
F3

59
0.
50
09

R
12

84
0.
06
69

R
H
R
11

10
0.
99
39

R
F1

1
35

0.
65
13

R
H
R
2

60
0.
50
09

R
13

85
0.
06
69

E
R
H
11

11
0.
99
20

R
F1

2
36

0.
65
06

R
FH

1
61

0.
50
09

R
H
9

86
0.
06
62

R
H
F6

12
0.
99
18

R
F5

37
0.
50
09

R
H
8

62
0.
47
34

R
H
F3

87
0.
04
19

E
R
H
10

13
0.
99
08

R
10

38
0.
50
09

E
R
H
1

63
0.
46
62

R
H
R
10

88
0.
04
08

E
R
H
9

14
0.
98
91

R
15

39
0.
50
09

R
H
R
1

64
0.
44
76

R
F6

89
0.
04
00

R
H
F5

15
0.
98
36

R
14

40
0.
50
09

R
H
5

65
0.
43
25

R
20

90
0.
03
99

E
R
H
14

16
0.
98
29

R
6

41
0.
50
09

R
F8

66
0.
40
37

R
H
R
9

91
0.
03
97

E
R
F1

2
17

0.
98
12

E
R
F2

42
0.
50
09

R
2

67
0.
39
98

R
18

92
0.
03
93

E
R
H
7

18
0.
98
10

R
3

43
0.
50
09

E
R
F3

68
0.
38
57

R
19

93
0.
03
88

E
R
F1

1
19

0.
97
99

R
H
4

44
0.
50
09

E
R
F1

69
0.
38
32

R
9

94
0.
03
84

E
R
H
8

20
0.
97
87

E
R
F4

45
0.
50
09

R
H
2

70
0.
36
30

R
H
R
5

95
0.
03
83

E
R
F1

0
21

0.
97
78

R
H
6

46
0.
50
09

R
H
7

71
0.
34
89

R
H
R
6

96
0.
03
83

E
R
H
13

22
0.
96
61

E
R
H
3

47
0.
50
09

R
H
1

72
0.
34
15

R
H
R
3

97
0.
03
75

R
H
F4

23
0.
92
67

R
F9

48
0.
50
09

R
25

73
0.
28
13

R
H
R
4

98
0.
03
74

R
H
R
8

24
0.
92
03

R
26

49
0.
50
09

R
H
F2

74
0.
27
54

R
F1

0
99

0.
03
50

E
R
F1

3
25

0.
90
67

R
1

50
0.
50
09

R
F4

75
0.
25
58

R
F7

10
0

0.
02
90

R
22

10
1

0.
02
79

R
21

10
2

0.
02
73

Table VII.
Route ranking results

727

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



it is necessary to perform an analysis beyond the scope of this paper, once the same storage
building may be used for stocking different products. Additional information regarding all
input variables is shown in Table X.

When analyzing outputs, the DMUs that have a slack equal to 0 for harvest and in-farm
storage capacity are 83 and 66, respectively. Neither Brazilian routes with rail
transportation nor Brazilian routes with waterway transportation presented slacks for
harvest at their final destination. For all the other DMUSs, it may be recommended to
increase harvest. No American routes with waterway transportation present slacks for
in-farm storage capacity. No Brazilian routes with exclusive transportation by barge or by
train present slack for the same variable. It may be recommended to increase in-farm
storage capacity for other routes; however, it is necessary to perform an analysis beyond the
scope of this paper, looking into the fact that the storage building may be used for stocking
different products.

In order to determine corridor efficiency, the arithmetic efficiency was calculated for all
the routes of each corridor. As can be seen from the results in Table XI, American corridors
are among the most efficient. Although Missouri presents the most efficient route, the most
efficient corridor, considering the total of routes, connectsMinnesota to Gulf of Mississippi.

Strongly efficient DMUs Weakly efficient DMUs
Code Route Code Route

EH1 New Madrid-Golfo do Mississipi ERF15 New Madrid-NPW
ERF10 Lac Qui Parle-NPW ERF2 Champaign-NPW
ERF11 Conttonwood-NPW ERF4 Plymouth-NPW
ERF12 Faribault-NPW ERF5 Woodbury-NPW
ERF13 Nodaway-NPW ERF6 Webster-NPW
ERF14 Audrain-NPW ERF9 Montgomery-NPW
ERF7 Knox-NPW ERH2 Champaign-Golfo do Mississipi
ERH10 Lac Qui Parle-Golfo do Mississipi ERH3 Mclean-Golfo do Mississipi
ERH11 Cottonwood-Golfo do Mississipi R1 Ijuí-Rio Grande
ERH12 Faribault-Golfo do Mississipi R10 Dourados-Paranaguá
ERH13 Nodaway-Golfo do Mississipi R11 Dourados-Santos
ERH14 Audrain-Golfo do Mississipi R14 Chapadão do Sul-Paranaguá
ERH4 Plymouth-Golfo do Mississipi R15 Chapadão do Sul-Santos
ERH5 Woodbury-Golfo do Mississipi R16 Jataí-Paranaguá
ERH6 Webster-Golfo do Mississipi R17 Jataí-Santos
ERH7 Knox-Golfo do Mississipi R26 Sorriso-Santos
ERH8 Clinton-Golfo do Mississipi R3 Bagé-Rio Grande
ERH9 Montgomery-Golfo do Mississipi R6 Guarapuava-Paranaguá
F1 Santa Maria-Rio Grande RF3 Dourados-Santos
F2 Londrina-Paranaguá RF5 Jataí-Paranaguá
F3 Guarapuava-Paranaguá RF9 Sorriso-Paranaguá
F4 Cascável-Paranaguá RFH1 Ijuí-Rio Grande
F5 Chapadão do Sul-Santos RH4 Ijuí-Rio Grande
FH1 Santa Maria-Rio Grande RH6 Bagé-Rio Grande
RF12 Sorriso-Santos RHR2 Jataí-Santos
RF2 Ijuí-Rio Grande
RHF1 Jataí-Santos
RHF4 Sorriso-Santos
RHF5 Canarana-Santos
RHF6 Primavera do Leste-Santos
RHR11 Primavera do Leste-Santos
RHR12 Primavera do Leste-Santos
RHR8 Sorriso-Santos

Table VIII.
Originally tied DMUs
divided between weak
and strong efficiency.
The horizontal bar
separates countries

728

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



5. Discussions
Two major obstacles were raised during this research. The first obstacle was to guarantee
that the estimation of fuel consumption for Brazil was realistic. Results were compared with
estimative of experts and proved to be adequate in an average scenario. The second obstacle
was to guarantee that the proposed variables were suitable to the model. The choice of
variables was based on literature review, data availability, and dimensions of sustainability.
The PCA analysis supported the proposal of the variables. Ranking results were similar and
coherent with literature.

DMU Benchmarks

ERF1 ERF2 ERF6 RF12 RF2 RHR12
ERF3 ERF6 ERH6 ERH9 RHR12
ERF8 ERF12 ERF15 ERF6 ERF9 ERH9 RHR12
ERH1 ERH2 ERH4 ERH6 ERH9 F3 RHR12
R12 EH1 ERF14 ERH12 ERH14
R13 EH1 ERF14 ERH12 ERH14
R18 EH1 ERF14 ERH12
R19 EH1 ERF14 ERH12
R2 EH1 ERH9 F3 RH6
R20 EH1 ERH12 ERH14 ERH9
R21 ERF15 RHF4 RHR12
R22 EH1 RHF4 RHR12
R23 ERH12 ERH4 RF2 RHR8
R24 ERH12 ERH4 ERH8 RF2 RHR8
R25 RHF4 RHR8
R4 ERH9 F3 RH4 RHR12
R5 ERH6 ERH9 F3 RF2
R7 ERH6 ERH9 F3 RF2
R8 ERH6 ERH8 ERH9 F3
R9 ERH4 ERH8 RF2 RHR8
RF1 ERH4 ERH9 F3 RF2
RF10 ERF12 ERH12 RHF4
RF11 ERF6 ERH12 ERH4 ERH9 RF12 RF2
RF4 EH1 ERF14 ERH12
RF6 EH1 ERF14 ERH12
RF7 ERF15 RHF4 RHR12
RF8 ERF6 ERH12 ERH4 RHF4
RH1 ERH10 ERH4 RF2 RHF4
RH2 ERH9 F3 RH6
RH3 ERH9 F3 RH6
RH5 EH1 ERH9 F3 RH6
RH7 ERH10 ERH4 RF2 RHF4
RH8 ERH4 ERH9 F3
RH9 ERH4 ERH9 F3
RHF2 EH1 ERF12 ERH12
RHF3 EH1 ERF14 ERH12
RHR1 RHF1 RHR2
RHR10 EH1 RHF6
RHR3 EH1 ERF14 ERH12
RHR4 EH1 ERF14 ERH12
RHR5 EH1 ERF14 ERH12
RHR6 EH1 ERF14 ERH12
RHR7 RHF4 RHR8
RHR9 ERF7 RHF4 RHF6

Table IX.
Benchmarks for

inefficient
routes (DMUs)

729

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Regarding results, Table XII summarizes the resemblance among the more and less efficient
routes. Among the efficient routes, the combination of two modes in the same route is
predominant. This is in accordance with the green premise, which prioritizes multimodal
corridors (Panagakos, 2016). According to a previous work (Oliveira and Cicolin, 2016), the
three most efficient routes in Brazil are combinations of two modes. Among the least
efficient, unimodal routes are more frequent, followed by tri-modal routes, suggesting that a

DMU
Fuel consumption

(%)
Planted area

(%)
Emissions

(%)
Disposal
factor (%)

Fatalities
(%)

Off-farm static storage
capacity (%)

ERF1 19.45 41.42 14.63 3.46
ERF3 9.11 34.33 8.04 1.96 2.28
ERF8 24.89 42.28 25.14
ERH1 7.23 20.98 14.28 5.88
R12 13.81 21.83 9.82 21.72
R13 13.81 21.83 9.84 21.72
R18 8.32 0.45 12.82 4.36 12.86
R19 8.35 0.25 12.92 4.85 13.08
R2 28.36 10.00 62.57 35.07 56.12
R20 8.19 12.93 4.18 13.11
R21 10.53 10.63 10.32 5.04 0.31
R22 10.87 10.95 10.65 4.98 0.91
R23 58.55 0.05 61.91 48.59 36.06
R24 60.57 65.07 45.25 37.74
R25 92.77 94.91 82.75 36.36
R4 18.59 34.49 22.69 43.42 2.02
R5 12.12 22.44 1.38 24.09 2.50
R7 10.73 18.67 1.10 19.71 1.02
R8 14.58 1.58 23.96 16.77 24.50
R9 7.97 0.74 11.11 1.75 9.99
RF1 4.15 11.76 6.14 16.91 2.55
RF10 12.40 12.96 12.16 4.22 2.94
RF11 32.20 27.30 29.10
RF4 9.08 0.55 18.78 11.12 16.79
RF6 7.01 0.61 12.29 5.89 12.15
RF7 10.69 10.88 10.61 3.58 0.86
RF8 58.04 0.58 61.07 55.03 29.49
RH1 40.45 0.62 71.49 3.85 65.63
RH2 19.75 12.11 31.31 24.26 29.42 2.21
RH3 14.87 19.78 22.46 10.08 20.28 3.56
RH5 26.58 9.54 63.85 48.52 54.63
RH7 43.00 1.33 57.13 8.30 14.90
RH8 16.73 10.80 21.07 21.01 16.19 1.00
RH9 12.40 18.64 17.93 8.88 14.46 2.66
RHF2 5.41 0.16 10.05 6.88 10.89
RHF3 6.59 0.13 11.16 6.33 10.80
RHR1 6.33 26.08 12.80 29.20
RHR10 3.72 0.38 5.80 5.06 4.04 0.38
RHR3 7.56 0.45 11.97 9.33 12.04
RHR4 7.55 0.40 11.99 9.39 12.14
RHR5 7.37 0.06 11.99 8.68 11.90
RHR6 7.36 0.01 12.01 8.74 11.95
RHR7 3.35 9.15 11.61 10.12
RHR9 3.79 5.52 5.20 3.84
Average 18.30 3.88 27.14 16.15 19.95 2.16

Table X.
Slack analysis of
input variables

730

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



Co
de

R
an
k

E
ff
ic
ie
nc
y

O
ri
gi
n

D
es
tin

at
io
n

Co
de

R
an
k

E
ff
ic
ie
nc
y

O
ri
gi
n

D
es
tin

at
io
n

E
C9

1
0.
99
33

M
N

G
ul
f
of

M
is
si
ss
ip
pi

E
C6

11
0.
50
09

IA
N
PW

E
C1

2
0.
98
72

M
O

G
ul
f
of

M
is
si
ss
ip
pi

E
C3

12
0.
48
93

IL
G
ul
f
of

M
is
si
ss
ip
pi

E
C7

3
0.
98
29

IN
G
ul
f
of

M
is
si
ss
ip
pi

C2
13

0.
44
67

PR
Pa

ra
na
gu

á
E
C1

0
4

0.
97
96

M
N

N
PW

E
C4

14
0.
42
32

IL
N
PW

E
C5

5
0.
80
95

IA
G
ul
f
of

M
is
si
ss
ip
pi

C4
15

0.
35
62

M
S

Pa
ra
na
gu

á
E
C2

6
0.
75
88

M
O

N
PW

C6
16

0.
27
03

G
O

Pa
ra
na
gu

á
C1

7
0.
68
78

R
S

R
io

G
ra
nd

e
C7

17
0.
24
04

G
O

Sa
nt
os

E
C8

8
0.
65
86

IN
N
PW

C8
18

0.
23
60

M
T

Pa
ra
na
gu

á
C5

9
0.
57
77

M
S

Sa
nt
os

C3
19

0.
06
36

PR
Sa
nt
os

C9
10

0.
57
48

M
T

Sa
nt
os

Table XI.
Corridor efficiencies

731

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



limit of modes to multimodal routes be considered efficient. Three or more modes in the
same route may induce inefficiency.

Road mode is highly placed among the more and the less efficient routes. In the USA,
trucks are used for transporting cargo through short distances (less than 200 km on
average), while, in Brazil, trucks may be used for transportation through more than
3,000 km. American routes are dominant among the most efficient, while there is no
American route among the least efficient, suggesting that short road routes are more
adequate to strengthen efficiency.

Among the efficient routes, trains (F1, F2, F3, F4, and F5) and barges (EH1, FH1) are
frequently present. In the USA, the combination of truck and barge is the most frequent
occurrence among efficient DMUs. Among the least efficient, long routes performed
exclusively by truck are the most common (15 occurrences), followed by the combination
of barges and trucks (ten occurrences). The barges are exclusively from the Paraná
Waterway, which is limited, by the infrastructural project, to a maximum combination of
three barges and tow. In the USA, combinations of 15 barges and a tow can navigate
through the Mississippi River. Also, the Paraná Waterway is not directly connected to
Port of Santos, so it is mandatory to perform the last kilometers of the route by truck;
however, this situation is does not occur in the USA. The results suggest that it is
recommended to focus investments on waterways projects without structural limitations
and directly connected to ports. Oliveira and Cicolin (2016) analyzed routes in Amazon
River (with a barge configuration similar to Mississippi, but still under-utilized) and
concluded that they were among the most efficient routes.

In accordance with the fact that the existence of barges or trains supports an efficient
route, Marquez and Cantillo (2013) analyzed the cost of Colombian and multimodal routes.
The authors concluded that the barges, followed by trains and trucks, present less external
costs. Nevertheless, a caveat is in order: the study did not consider the risks of drought and
did not specify the type of barge combination.

In 2007, the Brazilian Government began a comprehensive infrastructural
improvement strategy with major institutional and regulatory changes to facilitate
agricultural exports. The objective is that within 15-20 years, the railways’ participation
will increase from 25 to 35 percent; waterways from 13 to 29 percent; and truck
shipments will be reduced by 28 percent, from 58 to 30 percent. In January 2007, the
Brazilian Government created the Growth Acceleration Plan 1 (PAC 1) in order to modify
the transportation matrix. This plan promoted sustainable social and economic
development by generating employment, income and reducing regional inequalities.
By March 2010, the Government announced a second Growth Acceleration Plan (PAC 2)
2011-2014 (Salin, 2016).

The 2015 Transportation assessment report and the ninth evaluation results of Growth
Acceleration Program 2 (PAC 2), 2011-2015, showed that Brazil did not finish the projects as
planned. An example is highway BR-163 which began in PAC 1. The 619 miles highway,
connecting Brazil’s Midwest to the Amazon River, has not been completed. The completion
of this highway will significantly reduce transportation costs to the Amazon River ports
(Salin, 2016). The present study suggests investments in waterways without structural
limitations, such as Amazon-Tapajós, should be prioritized.

Number of modes Type of modes Country
One Two Three Road Rail Waterway USA Brazil

30 most efficient 6 17 7 23 15 19 16 14
30 least efficient 15 7 8 30 7 10 0 30

Table XII.
Summary of the
30 top and less
efficient routes

732

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



6. Conclusions
The DEA was presented as a solution tool for benchmarking the efficiency of freight
transport routes and corridors. The choice of variables was guided by the availability of the
data, the dimensions of sustainability, and literature review. Data analysis for maintaining
variables in the model was performed guided by the PCA.

The DEA is an adequate tool because it attributes the best weight possible to each
analyzed unit, overcoming the weight obstacles faced by a previous work (Panagakos,
2016). A few of the restrictions cited by Panagakos were: weights required for calculation
rely heavily on user interpretation, attributed weights can lead to misinterpretation, weights
change over time, and weights may not reflect specific characteristics of a country in
transport operations.

The weight choice with DEA is not dependent on human attributions and subject to
political or particular interpretations. As a matter of fact, weights change with time and
consequently, the data will also change. DEA applied with the updated data will automatically
apply updated weights after a certain period of time, resulting in the use of the best possible
attributes. Country-specific characteristics can be resolved with an accurate choice of
variables, validated by literature review, statistical tools, and stakeholder consulting.

The best DEA weight attribution may cause many ties. In the present application, the
problem of ties was overcome by the tiebreaking method of a composite index (Leta et al.,
2005). Hence, an initial method for benchmarking freight transport routes and corridors
was formulated. This method may be used to direct private and public investments on
logistics infrastructure. Corridor benchmarking is a rare topic in the literature. Previous
works normally focused on some specific and limited corridor performance characteristics,
such as cost. The main contribution of this research is that it expands the discussion about
corridor benchmarking and it focuses on efficiency as a whole. The fact that the method
used can be reapplied in different contexts in logistics is also a significant contribution.

Among DEA models, the SBM was chosen, because it minimizes inputs and
simultaneously maximizes outputs. Since SBM is not affected by units of measure, it is
possible to compare variables originated from different dimensions (economic, social, and
environmental). The results were similar to those of Oliveira and Cicolin (2016), which applied
the DEA-BCC model to measure Brazilian route efficiencies, and to a respected annual
international report (Salin, 2016) that compares soybean transportation in both countries.
The report also shows that routes in the Southern Brazilian regions (C1) are less onerous.

The practical implications show that American routes are more efficient than Brazilian
ones. Routes with terminals for exchanging transportation mode tend to be more efficient,
but there may be a limit for multimodality. The most efficient routes tend to present only
two modes: short truck trips and long barge trips or short truck trips and long train trips.
Inland waterway transportation tends to be efficient only with big combinations of barges
(i.e. 15 barges pushed by a tow) and when they are directly connected to the final destination
port. In accordance with a previous work that analyzes current Brazilian transport matrix
(Barros et al., 2015), the finding presented here not only contribute to the literature through
their empirical application, but may also assist Brazilian authorities in setting up more
adequate policies and investments to minimize the country’s excessive dependence on road
transportation. The results suggest that an increased focus should be given to railways and
inland waterways projects.

For further studies, the expansion and inclusion of variables such as maintenance route
costs, trade imbalance, lead time, type of cargo, and a variable that represents social values,
such as the quality of life of the operator, is recommended. It is also recommended that a risk
analysis considering environmental catastrophic events, such as droughts, be included.
Slack interpretation of SBM to variables in- and off-farm storage capacity may point the
number of storages that should be built as to increase route efficiency. Further studies

733

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)



should focus on guiding storage investments. It is also recommended to include link and
carryover variables, and to explore the possibilities of other DEA models, such as network
and dynamic models; structural window analysis; and other tiebreaking methods.

References

ABIPECAS – Brazilian Association of Automotive Part Industries (2011), “Survey of Brazilian
circulating fleet (Levantamento da frota circulante brasileira)”, Automotive Business Magazine,
p. 6, available at: www.automotivebusiness.com.br/pdf/pdf_125.pdf (accessed October 31, 2016).

Adler, N. and Golany, B. (2007), “PCA-DEA”, in Zhu, J. and Cook, W.D. (Eds),Modeling Data Irregularities
and Structural Complexities in Data Envelopment Analysis, Springer, New York, NY, pp. 139-166.

Adler, N. and Yazhemsky, E. (2010), “Improving discrimination in data envelopment analysis: PCA-DEA
or variable reduction”, European Journal of Operational Research, Vol. 202 No. 1, pp. 273-284.

AHRANA – Administration of Paraná Waterways (2005), “The Paraná Waterways (A Hidrovia do
Paraná)”, available at: www.cooperhidro.com.br/palestras/ruy-ahrana.pdf (accessed
October 31, 2016).

AHRANA – Administration of Paraná Waterways (2012), “Paraná Waterways data and information
(Dados e Informações Hidrovia do Rio Paraná)”, available at: www.ahrana.gov.br/dados_
informacoes.html (accessed October 31, 2016).

AHRANA – Administration of Paraná Waterways (2013), “Load handling statics (Estatísticas de
Movimentação de Carga)”, available at: www.ahrana.gov.br/dados_operacionais.html (accessed
October 31, 2016).

Alianca – Trevisa Investments (2016), “Fleet (Frota)”, available at: www.trevisa.com.br/alianca/frota.
html (accessed October 31, 2016).

ALL – Latin America Logistics (2016), “Rail network (Malha Ferroviária)”, available at: http://pt.
rumolog.com/default_pti.asp?idioma=0&conta=45 (accessed October 31, 2016).

Arnold, J. (2006), “Best practices in Management of International Trade Corridors, Trade Logistics
Group, Transport Papers TP-13”, The World Bank Group, Washington, DC, available at: www.
mcli.co.za/mcli-web/downloads/docs/Best_Practices_in_Management_of_International_Trade_
Corridors.pdf (accessed October 31, 2016).

Banker, R.D., Charnes, A. and Cooper, W.W. (1984), “Some models for estimating technical and scale
inefficiencies in data envelopment analysis”, Management Science, Vol. 30 No. 9, pp. 1078-1092.

Barros, C.P., Gil-Alana, L.A. and Wanke, P. (2015), “An empirical analysis of freight transport
traffic modes in Brazil, 1996-2012”, Transportation Planning and Technology, Vol. 38 No. 13,
pp. 305-319.

Baumel, P. (2011), “Measuring bulk product transportation fuel efficiency”, Transportation Research
Forum, Washington, DC, pp. 79-88, available at: www.trforum.org/journal/downloads/2011v50
n1_05_FuelEfficiency.pdf (accessed October 31, 2016).

Baumel, P., Hurburgh, C.R. and Lee, T. (2015), “Estimates of total fuel consumption in transporting grain
from Iowa to major grain countries by alternatives modes and routes”, Iowa Grain Quality
Initiative, available at: www.extension.iastate.edu/grain/topics/EstimatesofTotalFuelConsumption.
htm (accessed October 31, 2016).

Bureau of Transportation Statistics (2016), “National Transportation and Statistics”, available at:
www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/
html/table_01_26.html_mfd (accessed October 31, 2016).

Camioto, F.C., Mariano, E.B. and Rebelatto, D.A.N. (2014), “Efficiency in Brazil’s industrial sectors in
terms of energy and sustainable development”, Environmental Science & Policy, Vol. 37, March,
pp. 50-60, doi: https://doi.org/10.1016/j.envsci.2013.08.007.

Castro, V.F. and Frazzon, E.M. (2017), “Benchmarking of best practices: overview of academic
literature”, Benchmarking: An International Journal, Vol. 24 No. 3, pp. 617-634.

734

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)

www.automotivebusiness.com.br/pdf/pdf_125.pdf
www.cooperhidro.com.br/palestras/ruy-ahrana.pdf
www.ahrana.gov.br/dados_informacoes.html
www.ahrana.gov.br/dados_informacoes.html
www.ahrana.gov.br/dados_operacionais.html
www.trevisa.com.br/alianca/frota.html
www.trevisa.com.br/alianca/frota.html
http://pt.rumolog.com/default_pti.asp?idioma=0&#x00026;conta=45
http://pt.rumolog.com/default_pti.asp?idioma=0&#x00026;conta=45
www.mcli.co.za/mcli-web/downloads/docs/Best_Practices_in_Management_of_International_Trade_Corridors.pdf
www.mcli.co.za/mcli-web/downloads/docs/Best_Practices_in_Management_of_International_Trade_Corridors.pdf
www.mcli.co.za/mcli-web/downloads/docs/Best_Practices_in_Management_of_International_Trade_Corridors.pdf
www.trforum.org/journal/downloads/2011v50n1_05_FuelEfficiency.pdf
www.trforum.org/journal/downloads/2011v50n1_05_FuelEfficiency.pdf
www.extension.iastate.edu/grain/topics/EstimatesofTotalFuelConsumption.htm
www.extension.iastate.edu/grain/topics/EstimatesofTotalFuelConsumption.htm
www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/html/table_01_26.html_mfd
www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/html/table_01_26.html_mfd
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.envsci.2013.08.007&isi=000333723800005&citationId=p_15
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&system=10.1108%2FBIJ-03-2016-0031&citationId=p_16
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1007%2F978-0-387-71607-7_8&citationId=p_2
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1007%2F978-0-387-71607-7_8&citationId=p_2
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.ejor.2009.03.050&isi=000271700800032&citationId=p_3
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1287%2Fmnsc.30.9.1078&isi=A1984TJ33900004&citationId=p_10
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1080%2F03081060.2014.997452&citationId=p_11


Charnes, A., Cooper, W.W. and Rhodes, E. (1978), “Measuring the efficiency of the decision making
units”, European Journal Of Operational Research, Vol. 6 No. 2, pp. 429-444.

Charnes, A., Cooper, W.W., Rousseau, J.J. and Semple, J. (1987), “Data envelopment analysis and
axiomatic notions of efficiency and reference sets”, CCS Research Report No. 558, Graduate
School of Business, Center for Cybernetic Studies, University of Texas, Austin, TX.

Charnes, A., Cooper, W.W., Golany, B., Seiford, L. and Stutz, J. (1985), “Foundations of data
envelopment analysis for Pareto-Koopmans efficient empirical production functions”, Journal of
Econometrics, Vol. 30 Nos 1-2, pp. 91-107.

CONAB – National Supply Company (2005), “Agricultural Storage in Brazil”, Armazenagem agrícola
no Brasil, available at: www.conab.gov.br/OlalaCMS/uploads/arquivos/7420aabad20
1bf8d9838f446e17c1ed5..pdf (accessed October 31, 2016).

CONAB – National Supply Company (2014), “Historical series” (Séries Históricas)”, available at: www.
conab.gov.br/conteudos.php?a=1252&t=&Pagina_objcmsconteudos=3#A_objcmsconteudos
(accessed October 31, 2016).

CONTRAN – National Traffic Counsil (1999), “Law no 7.408, from 1985”, Resolution 104 from
December 21, 1999, “2. Legal Limits” (2. Limites Legais), available at: www1.dnit.gov.br/
Pesagem/qfv pdf.pdf (accessed October 31, 2016).

Cook, W.D., Tone, K. and Zhu, J. (2014), “Data envelopment analysis: prior to choosing a model”,
Omega, Vol. 44, April, pp. 1-4, available at: https://doi.org/10.1016/j.omega.2013.09.004

Dubey, R., Gunasekaran, A. and Papadopoulos, T. (2017), “Green supply chain management: theoretical
framework and further research directions”, Benchmarking: An International Journal, Vol. 24
No. 1, pp. 184-218.

EMATER – State Company of Technical Assistance and Rural Extension (2014), “Follow-up of the
harvest: 2014/2015 harvest (Acompanhamento da Safra: Safra 2014/2015)”, (Séries Históricas),
EMATER, Porto Alegre, p. 6, available at: www.emater.tche.br/site/servicos/informacoes-
agropecuarias.php#acompanhamento-de-safra (accessed February 5, 2018).

Ferroeste – Secretariat of Infrastructure and Logistics (2016), “Rail network (Malha ferroviária)”,
available at: www.ferroeste.pr.gov.br/modules/conteudo/conteudo.php?conteudo=47 (accessed
February 5, 2018).

Greenhouse Gas Protocol – World Resources Institute (2016), “Calculating CO2 emissions from mobile
sources”, Greenhouse Gas Protocol, Washington, DC, p. 11, available at: www.ghgprotocol.org/
files/ghgp/tools/co2-mobile.pdf (accessed June 17, 2016).

Grigoroudis, E., Petridis, P. and Arabatzis, G. (2014), “RDEA: a recursive DEA based on algorithm for
the optimal design of biomass supply chain network”, Renewable Energy, Vol. 71, November,
pp. 113-122, available at: https://doi.org/10.1016/j.renene.2014.05.001

Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W. (1995),Multivariate Data Analysis, Prentice-Hall,
Englewood Cliffs. NJ.

IBGE – Brazilian Institute of Geography and Statistics (2016a), “Planning Ministry (Ministério do
Planejamento)”, “Regional division” (Divisão Regional), available at: www.ibge.gov.br/home/
geociencias/geografia/default_div_int.shtm?c=1 (accessed February 5, 2018).

IBGE – Brazilian Institute of Geography and Statistics (2016b), “Planning Ministery (Ministério do
Planejamento)”, Aggregated Database (Banco de Dados Agregados), available at: www.sidra.
ibge.gov.br/bda/ (accessed August 30, 2016).

IMBEES – Institute Mauro Borges of Statics and Socioeconomic Studies (2015), “Secretariat of State of
Goiás of Management and Planning (Secretaria de Estado de Goiás de Gestão e Planejamento)”,
Municipal Statistics: Historical Series (Estatísticas Municipais: Séries Históricas), available at:
www.imb.go.gov.br/ (accessed February 5, 2018).

IMEA – Institute of State of Mato Grosso of Agricultural Economics (2015), “Weekly Soy Newsletter
(Boletim Semanal de Soja)”, 358. ed., IMEA, Cuiabá, p. 12, available at: www.imea.com.br/
upload/publicacoes/arquivos/R404_2015_06_19_BSSoja.pdf (accessed February 5, 2018).

735

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)

www.conab.gov.br/OlalaCMS/uploads/arquivos/7420aabad201bf8d9838f446e17c1ed5..pdf
www.conab.gov.br/OlalaCMS/uploads/arquivos/7420aabad201bf8d9838f446e17c1ed5..pdf
www.conab.gov.br/conteudos.php?a=1252&#x00026;t=&#x00026;Pagina_objcmsconteudos=3#A_objcmsconteudos
www.conab.gov.br/conteudos.php?a=1252&#x00026;t=&#x00026;Pagina_objcmsconteudos=3#A_objcmsconteudos
www1.dnit.gov.br/Pesagem/qfv pdf.pdf
www1.dnit.gov.br/Pesagem/qfv pdf.pdf
https://doi.org/10.1016/j.omega.2013.09.004
www.emater.tche.br/site/servicos/informacoes-agropecuarias.php#acompanhamento-de-safra
www.emater.tche.br/site/servicos/informacoes-agropecuarias.php#acompanhamento-de-safra
www.ferroeste.pr.gov.br/modules/conteudo/conteudo.php?conteudo=47
www.ghgprotocol.org/files/ghgp/tools/co2-mobile.pdf
www.ghgprotocol.org/files/ghgp/tools/co2-mobile.pdf
https://doi.org/10.1016/j.renene.2014.05.001
www.ibge.gov.br/home/geociencias/geografia/default_div_int.shtm?c=1
www.ibge.gov.br/home/geociencias/geografia/default_div_int.shtm?c=1
www.sidra.ibge.gov.br/bda/
www.sidra.ibge.gov.br/bda/
www.imb.go.gov.br/
www.imea.com.br/upload/publicacoes/arquivos/R404_2015_06_19_BSSoja.pdf
www.imea.com.br/upload/publicacoes/arquivos/R404_2015_06_19_BSSoja.pdf
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2F0304-4076%2885%2990133-2&isi=A1985AZA7200006&citationId=p_19
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2F0304-4076%2885%2990133-2&isi=A1985AZA7200006&citationId=p_19
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.omega.2013.09.004&isi=000331431600001&citationId=p_23
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&system=10.1108%2FBIJ-01-2016-0011&isi=000395590900010&citationId=p_24
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.renene.2014.05.001&isi=000340976600015&citationId=p_28
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2F0377-2217%2878%2990138-8&citationId=p_17


IMTT – Institute of Mobility and Land Transport (2011), “Rail transportation: annual safety report of
2011 (Transporte Ferroviário Relatório Anual de Segurança de 2011)”, IMTT, São Paulo, p. 78,
available at: www.imt-ip.pt/sites/IMTT/Portugues/IMTT/relatoriosectoriais/Documents/
RelatorioAnualSegurancaTranspFerroviario2011.pdf (accessed February 5, 2018).

Kengpol, A., Tuammee, S. and Tuominen, M. (2014), “The development of a framework for route
selection in multimodal transportation”, The International Journal of Logistics Management,
Vol. 25 No. 3, pp. 581-610, available at: www.emeraldinsight.com/doi/abs/10.1108/IJLM-05-20
13-0064 (accessed June 2016).

Leta, F.R., Mello, J.C.C.B.S., Gomes, E.G. and Meza, L.A. (2005), “Métodos De Melhora De Ordenação em
DEA Aplicados à Avaliação Estática de Tornos Mecânicos”, Investigação Operacional, Vol. 25
No. 2, pp. 229-242, available at: www.researchgate.net/publication/228375009_Metodos_de_
melhora_de_ordenacao_em_DEA_aplicados_a_avaliacao_estatica_de_tornos_mecanicos
(accessed June 2016).

Macrologística (2011), “CNI – National Confederation of Industry (Confederação Nacional da
Indústria)”, Competitive Midwest Project (Projeto Centro-Oeste Competitivo), São Paulo, p. 75,
available at: http://arquivos.portaldaindustria.com.br/app/conteudo_18/2015/10/22/9958/
ProjetoCentro-OesteCompetitivo.pdf (accessed February 5, 2018).

Macrologística (2013), “CNI – National Confederation of Industry (Confederação Nacional da
Indústria)”, Competitve South Project (Projeto Sul Competitivo), São Paulo, p. 75, available at:
www.macrologistica.com.br/index.php/pt/midia/palestras-e-relatorios/96-projeto-sul-competitivo
(accessed February 5, 2018).

Maia, G.B.S., Pinto, A.R., Marques, C.Y.T., Lyra, D.D. and Roitman, F.B. (2013), “Overview of the
storage of agricultural products in Brazil (Panorama da armazenagem de produtos agrícolas no
Brasil)”, Magazine of BNDES, Vol. 1 No. 40, pp. 161-194, available at: www.bndes.gov.br/
SiteBNDES/export/sites/default/bndes_pt/Galerias/Arquivos/conhecimento/revista/rev4005.pdf
(accessed August 2016).

Mariano, E.B. and Rebelatto, D.A.N. (2014), “Transformation of wealth produced into quality of life:
analysis of the social efficiency of nation-states with the DEA’s triple index approach”, Journal
of The Operational Research Society, Vol. 65 No. 11, pp. 1664-1681, available at: www.palgrave-
journals.com/jors/journal/v65/n11/abs/jors2013132a.html (accessed June 2016).

Marquez, L. and Cantillo, V. (2013), “Evaluating strategic freight transport corridors including external
costs”, Transportation Planning and Technology, Vol. 36 No. 6, pp. 529-546, available at: www.
tandfonline.com/doi/abs/10.1080/03081060.2013.830892 (accessed June 2016).

Matsuda, T. and Goldsmith, P.D. (2009), “World soybean production: area harvested, yield, and
long-term projections”, International Food and Agribusiness Management Review, Vol. 12 No. 4,
pp. 143-162, available at: http://ageconsearch.umn.edu/bitstream/92573/2/20091023_Formatted.
pdf (accessed August 2016).

MDIC – Ministry of Industry, Foreign Trade and Services (2016), “Secretariat of Foreign Trade:
Aliceweb (Secretaria do Comério Exterior”, available at: http://aliceweb.mdic.gov.br/ (accessed
February 5, 2018).

Murray, D. (2016), “Barge fleet holding steady, getting younger”, The Waterway Journal Weekly, Vol. 129
No. 9, p. 1, available at: http://waterwaysjournal.net/Magazine/ThisWeeksTopNews/
BargeFleetHoldingSteady,GettingYounger.aspx (accessed February 5, 2018).

National Agricultural Statistics Services (NASS) (2016), “Statistics by state”, available at: www.nass.
usda.gov/Statistics_by_State/ (accessed June 2016).

National Highway Traffic Safety Administration (2016), “State Traffic Safety Info”, available at:
www-nrd.nhtsa.dot.gov/departments/nrd-30/ncsa/STSI/USAWEBREPORT.HTM (accessed
February 5, 2018).

Neves, M. (2012), “Total fleet of Brazilian locomotives grows by 126 units: market study (Frota total de
locomotivas brasileiras cresce em 126 unidades: Estudos de Mercado)”, Rail Magazine (Revista
Ferroviária), Vol. 62 No. 1, pp. 56-58, available at: www.revistaferroviaria.com.br/upload/Todas
as locomotivas 2011.pdf (accessed August 2016).

736

BIJ
25,2

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)

www.imt-ip.pt/sites/IMTT/Portugues/IMTT/relatoriosectoriais/Documents/RelatorioAnualSegurancaTranspFerroviario2011.pdf
www.imt-ip.pt/sites/IMTT/Portugues/IMTT/relatoriosectoriais/Documents/RelatorioAnualSegurancaTranspFerroviario2011.pdf
www.emeraldinsight.com/doi/abs/10.1108/IJLM-05-2013-0064
www.emeraldinsight.com/doi/abs/10.1108/IJLM-05-2013-0064
www.researchgate.net/publication/228375009_Metodos_de_melhora_de_ordenacao_em_DEA_aplicados_a_avaliacao_estatica_de_tornos_mecanicos
www.researchgate.net/publication/228375009_Metodos_de_melhora_de_ordenacao_em_DEA_aplicados_a_avaliacao_estatica_de_tornos_mecanicos
http://arquivos.portaldaindustria.com.br/app/conteudo_18/2015/10/22/9958/ProjetoCentro-OesteCompetitivo.pdf
http://arquivos.portaldaindustria.com.br/app/conteudo_18/2015/10/22/9958/ProjetoCentro-OesteCompetitivo.pdf
www.macrologistica.com.br/index.php/pt/midia/palestras-e-relatorios/96-projeto-sul-competitivo
www.bndes.gov.br/SiteBNDES/export/sites/default/bndes_pt/Galerias/Arquivos/conhecimento/revista/rev4005.pdf
www.bndes.gov.br/SiteBNDES/export/sites/default/bndes_pt/Galerias/Arquivos/conhecimento/revista/rev4005.pdf
www.palgrave-journals.com/jors/journal/v65/n11/abs/jors2013132a.html
www.palgrave-journals.com/jors/journal/v65/n11/abs/jors2013132a.html
www.tandfonline.com/doi/abs/10.1080/03081060.2013.830892
www.tandfonline.com/doi/abs/10.1080/03081060.2013.830892
http://ageconsearch.umn.edu/bitstream/92573/2/20091023_Formatted.pdf
http://ageconsearch.umn.edu/bitstream/92573/2/20091023_Formatted.pdf
http://aliceweb.mdic.gov.br/
http://waterwaysjournal.net/Magazine/ThisWeeksTopNews/BargeFleetHoldingSteady,GettingYounger.aspx
http://waterwaysjournal.net/Magazine/ThisWeeksTopNews/BargeFleetHoldingSteady,GettingYounger.aspx
www.nass.usda.gov/Statistics_by_State/
www.nass.usda.gov/Statistics_by_State/
www-nrd.nhtsa.dot.gov/departments/nrd-30/ncsa/STSI/USAWEBREPORT.HTM
www.revistaferroviaria.com.br/upload/Todas as locomotivas 2011.pdf
www.revistaferroviaria.com.br/upload/Todas as locomotivas 2011.pdf
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&isi=000208146300007&citationId=p_42
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&system=10.1108%2FIJLM-05-2013-0064&isi=000348438800007&citationId=p_35
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1057%2Fjors.2013.132&isi=000344084100004&citationId=p_40
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1057%2Fjors.2013.132&isi=000344084100004&citationId=p_40
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1080%2F03081060.2013.830892&isi=000324086400004&citationId=p_41


Ojima, A.L.R.O. (2004), “Analysis of logistics and competitiveness of Brazilian soy: an application of
spatial equilibrium model of quadratic programming (Análise da Movimentação Logística e
Competitividade da Soja Brasileira: Uma Aplicação de um Modelo de Equilíbrio Espacial de
Programação Quadrática)”, dissertation, University of Campinas, Campinas, available at: www.
bibliotecadigital.unicamp.br/document/?code=vtls000317206 (accessed June 2016).

Oliveira, A.L.R. and Cicolin, L.O.M. (2016), “Evaluating the logistics performance of Brazil’s corn exports:
a proposal of indicators”, African Journal of Agricultural Research, Vol. 11 No. 8, pp. 693-700.

Panagakos, G. (2016), “Green corridors basics”, in Psaraftis, H.N. (Ed.), Green Transportation Logistics:
The Quest for Win-Win Solution (International Series in Operations Research & Management
Science), Chapter 3, International Publishing Switzerland, Cham, pp. 81-121, available at: www.
springer.com/gp/book/9783319171746 (accessed June 2016).

Rocha, C.B. and Parré, J.L. (2009), “Study of spatial distribution of the agricultural sector of Rio Grande
do Sul (Estudo da Distribuição Espacial do Setor Agropecuário do Rio Grande do Sul)”,
Economic Analysis (Análise Econômica), Vol. 27 No. 52, pp. 139-160, available at: http://seer.
ufrgs.br/AnaliseEconomica/article/view/5159 (accessed June 2016).

Saen, R.F. (2005), “Developing a nondiscretionary model of slacks-based measure in data envelopment
analysis”, Applied Mathematics and Computation, Vol. 169 No. 2, pp. 1440-1447, available at:
https://doi.org/10.1016/j.amc.2004.10.053

Salin, D. (2016), “Soybean transportation guide: Brazil”, US Department of Agriculture, Agricultural
Marketing Service, Washington, DC, available at: www.ams.usda.gov/sites/default/files/media/
Brazil%20Soybean%20Transportation%20Guide%202015.pdf (accessed October 2016).

SEAB – Secretariat of Agriculture and Supply (2015), “Harvest Estimates (Estimativas de Safra)”,
available at: www.agricultura.pr.gov.br/ (accessed February 5, 2018).

Shao, Y. and Sun, C. (2016), “Performance evaluation of China’s air routes on network data envelopment
analysis approach”, Journal of Air Transport Management, Vol. 55, August, pp. 67-75, available at:
https://doi.org/10.1016/j.jairtraman.2016.01.006

SIGA – Agribusiness Geographic Information System (2015), “Technical circular: follow-up of the crop
2013/2014 (Circular Técnica: Acompanhamento da Safra 2013/2014)”, 56 ed., SIGA, Campo
Grande, p. 9, available at: http://sigaweb.aprosojams.org.br/ (accessed June 2016).

Simoes, A.J.C. and Hidalgo, C.A. (2011), “The Economic Complexity Observatory: an analytical tool for
understanding the dynamics of economic development”, Workshops at the Twenty-Fifth AAAI
Conference on Artificial Intelligence, San Francisco, CA, available at: http://atlas.media.mit.edu/
pt/resources/permissions/ (accessed June 2016).

Soy Transportation Coalition (2016), “Fuel efficiency: barge vs railroad vs truck”, available at: www.
soytransportation.org/Stats/Modal_FuelEfficiency.pdf (accessed February 5, 2018).

Texas Transportation Institute (TTI) (2007), “A modal comparison of domestic freight transportation
effects on the general public: final report”, National Waterway Foundation, Houston, TX, p. 69,
available at: www.marad.dot.gov/wp-content/uploads/pdf/Phase_II_Report_Final_121907.pdf
(accessed June 2016).

Texas Transportation Institute (TTI) (2012), “A modal comparison of domestic freight transportation
effects on the general public: final report”, 2nd ed., National Waterway Foundation, Houston,
p. 59, available at: www.nationalwaterwaysfoundation.org/study/FinalReportTTI.pdf (accessed
June 2016).

Tone, K. (2001), “A slacks-based measure of efficiency in data envelopment analysis”, European
Journal of Operational Research, Vol. 130 No. 3, pp. 498-509, available at: www.sciencedirect.
com/science/article/pii/S0377221799004075 (accessed June 2016).

Tongzon, J. (2001), “Efficiency measurement of selected Australian and other international ports using
data envelopment analysis”, Transportation Research Part A, Vol. 37 No. 2, pp. 107-122,
available at: https://doi.org/10.1016/S0965-8564(99)00049-X

US Corps of Engineers, US Army (2016), “Planning Center of Expertise for Inland Navigation (PCXIN)”,
available at: http://outreach.lrh.usace.army.mil/ (accessed February 5, 2018).

737

Data
envelopment

analysis

D
ow

nl
oa

de
d 

by
 U

N
E

SP
 A

t 1
1:

56
 2

5 
A

pr
il 

20
19

 (
PT

)

www.bibliotecadigital.unicamp.br/document/?code=vtls000317206
www.bibliotecadigital.unicamp.br/document/?code=vtls000317206
www.springer.com/gp/book/9783319171746
www.springer.com/gp/book/9783319171746
http://seer.ufrgs.br/AnaliseEconomica/article/view/5159
http://seer.ufrgs.br/AnaliseEconomica/article/view/5159
https://doi.org/10.1016/j.amc.2004.10.053
www.ams.usda.gov/sites/default/files/media/Brazil%20Soybean%20Transportation%20Guide%202015.pdf
www.ams.usda.gov/sites/default/files/media/Brazil%20Soybean%20Transportation%20Guide%202015.pdf
www.agricultura.pr.gov.br/
https://doi.org/10.1016/j.jairtraman.2016.01.006
http://sigaweb.aprosojams.org.br/
http://atlas.media.mit.edu/pt/resources/permissions/
http://atlas.media.mit.edu/pt/resources/permissions/
www.soytransportation.org/Stats/Modal_FuelEfficiency.pdf
www.soytransportation.org/Stats/Modal_FuelEfficiency.pdf
www.marad.dot.gov/wp-content/uploads/pdf/Phase_II_Report_Final_121907.pdf
www.nationalwaterwaysfoundation.org/study/FinalReportTTI.pdf
www.sciencedirect.com/science/article/pii/S0377221799004075
www.sciencedirect.com/science/article/pii/S0377221799004075
https://doi.org/10.1016/S0965-8564(99)00049-X
http://outreach.lrh.usace.army.mil/
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.5897%2FAJAR2015.10653&citationId=p_49
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2FS0377-2217%2899%2900407-5&isi=000167394300004&citationId=p_61
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2FS0377-2217%2899%2900407-5&isi=000167394300004&citationId=p_61
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1007%2F978-3-319-17175-3_3&citationId=p_50
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1007%2F978-3-319-17175-3_3&citationId=p_50
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1007%2F978-3-319-17175-3_3&citationId=p_50
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.jairtraman.2016.01.006&isi=000380594500007&citationId=p_55
https://www.emeraldinsight.com/action/showLinks?doi=10.1108%2FBIJ-11-2016-0175&crossref=10.1016%2Fj.amc.2004.10.053&isi=000232811600056&citationId=p_52


Vieira, N.M. (2002), “Caracterization of soy production chain in Goiás (Caracterização da Cadeia
Produtiva da Soja em Goiás) dissertation”, Federal University of Santa Catarina, Florianópolis,
available at: https://repositorio.ufsc.br/handle/123456789/83611 (accessed June 2016).

Waterways Council (2016), “Waterways systems”, available at: http://waterwayscouncil.org/
waterways-system/ (accessed February 5, 2018).

Further reading

ANTAQ – National Agency of Waterway Transportation (2014), “Statistical waterways yearly
handbook: port handling (Anuário Estatístico Aquaviário: Movimentação Portuária)”, available
at: www.antaq.gov.br/anuario/ (accessed October 31, 2016).

Casavant, K., Denicoff, M.R., Jessup, E., Taylor, A., Nibarger, D., Sear, D., Khachatryan, H., Mccracken, V.,
Prater, N., Marvin., J.E., Oleary, N., Marathon, N., Mcgregor, B., Olowolayemo, S and Blanton, B.
(2010), “Study of Rural Transportation Issues”, Research Reports No. 147544, United States
Department of Agriculture, Agricultural Marketing Service, Transportation and Marketing
Program, Washington, DC, available at: http://ntl.bts.gov/lib/32000/32800/32855/STELPRDC50
84108.pdf (accessed October 31, 2016).

DNIT – National Department of Transport Infrastructure (2009), “Statistical yearbook of federal
highways (Anuário Estatístico das Rodovias Federais)”, available at: www.dnit.gov.br/
download/rodovias/operacoes-rodoviarias/estatisticas-de-acidentes/anuario-2009.pdf (accessed
February 5, 2018).

FRA – Federal Railway Administration, Office of Safety Analysis (2010), “Total accidents/incidents
January-December (2010)”, available at: http://safetydata.fra.dot.gov/officeofsafety/publicsite/
summary.aspx (accessed February 5, 2018).

Ferreira, A. N. (2000), “Study of the effect of accidents on Tietê-Paraná waterway: preventive aspects
(Estudo do Efeito de Acidentes na Hidrovia Tietê-Paraná: Aspectos Preventivos)”,
dissertationPolytechnic School of University of São Paulo, São Paulo, available at: www.teses.
usp.br/teses/disponiveis/3/3135/tde-19042002-081248/pt-br.php (accessed June 2016).

NOAA – National Oceanic and Atmospheric Administration (2012), “Distances between United States
ports”, 12th ed., NOAA, Washington, p. 56, available at: www.nauticalcharts.noaa.gov/nsd/
distances-ports/distances.pdf (accessed June 2016).

Sheth, C., Triantis, K. and Teodorovic, D. (2007), “Performance evaluation of bus routes: a provider and
passenger perspective”, Transportation Research Part E: Logistics and Transportation Review,
Vol. 43 No. 4, pp. 453-478.

STATISTA – The Statistical Portal (2016), “Average age of freight rail cars in the United States from
2005 to 2011 (in years)”, available at: www.statista.com/statistics/245305/age-of-us-freight-rail-
cars/ (accessed February 5, 2018).

United States Department of Agriculture (USDA) (2012), “United Soybean Board Farm to market:
a soybean’s journey from field to costumer”, Informa Economics, available at: https://
unitedsoybean.org/wp-content/uploads/FarmToMarketStudy.pdf (accessed February 5, 2018).

Corresponding au