Prioritizing agricultural research projects with emphasis on
analytic hierarchy process (AHP)
M. Mortazavi, L. Ghanbari, M. Rajabbeigi and H. Mokhtari
Institute of Technical and Vocational High Education of Agriculture, P.O. Box 13145-1757, Tehran,
Iran; [email protected]
Abstract
This article reviews various methods of setting agricultural research priorities and introduces
application of Analytic Hierarchy Process (AHP) as a well-known method on Multi-Attribute
Decision Making (MADM) in prioritizing agricultural research. In addition, it illustrates how
to combine qualitative and quantitative attributes in setting priorities for agricultural research
programs and projects (sugar beet crop). For this purpose, final attributes of decision making are
initially identified and then, a decision hierarchy tree is drawn. Required data are made by means
of a paired comparison questionnaire through a participation of elites and researchers, followed by
estimations with an application of Expert Choice software. According to the results produced, the
research priorities consist of sustainable development, scientific development, research feasibility,
and economic development attributes with different weigh coefficients, respectively. Most priorities
are also accrued by a number of sub-attributes including conservation of genetic reserves, protection
of base resources, achieving knowledge and technology, and achieving new products and services.
Keywords: multi-attribute decision making (MADM), priority setting
Introduction
Recent century has witnessed an economy system based on science and technology emerging in
global relations as a new phenomenon, so that an economy is measured in terms of its nature and
potential through knowledge-orientation. Thus, many countries tend to pay specific attentions to
making investments in their national research systems, and in the agricultural studies in particular;
so that until mid-ninetieth, annual expenditures on agricultural research and development in various
countries were totally estimated about $ 33.2 billions of which developing countries shared $ 12.2
billions (Pardey, 1998).
Activities and efforts within an agricultural research system would lead to success when the system
may respond to issues, problems and expectations of operators and other beneficiary groups.
However, one should bear in mind that their wholly comprehensive responding is not feasible
in effect while these and other demands in agriculture sector appear to be so broad, diversified,
numerous and complicated as well as there are certain limitations in time, facilities and equipments,
financial and monitory resources, and human force. Therefore, in search for a wise and rational
remedy, there is no way out unless resources and facilities are optimally allocated to research
priorities. Hence, it is obvious that setting research priorities within agricultural research system
would be a major concern and challenge.
To optimize decision making processes in agricultural research management, it is preferred to
initially provide a comprehensive, explicit and updated database of research programs and projects,
in which programs and projects are determined in respect to either needs and demands of operators
and beneficiaries or the issues, problems and constraints to be studied.
EFITA conference ’09
31
Secondly, since it is impossible to undertake all researches simultaneously, due to existing
limitations, there should be prepared a preliminary list of the potential subjects to be studied
known as priority setting options according to the information gathered from the above database.
Thirdly, the potential impacts and results of each research option have to be estimated, assuming
their conduction and implementation of the results. Then, as the fourth step, existing options are
to be prioritized by means of an appropriate method. In the fifth step, according to the results from
priority setting, guidelines to include priorities into the research planning should be formulated.
Assessing options and setting their priorities are influenced by factors like indexes of decision
and key decision makers’ viewpoints as well as inclusion of organizational conditions, implying a
sort of complicated decision making. These indexes might be in their nature considered as either
quantitative or qualitative and/or both types of indexes, indicating the complexity of making
decisions on them, particularly when options are assessed to be favorable by some indexes while
being unfavorable by some others. In addition, since such decisions are often made in a group,
it is of a great challenge to combine views so that it would lead to a decision with the agreement
and consent of all the group members. Such a decision making environment tends to conform to
capabilities of Multi-Attribute Decision Making (MADM) method.
Most common methods of setting agricultural research priority
Many methods of priority setting for agricultural research have been used the most common of
which include: Mathematical Models, Scoring, Checklist, Rule of Thumb, Domestic Resource
Cost, economic Surplus, Cost-Benefit, and Simulation (Braunschweig, 2000).
Among the above-mentioned methods, ‘scoring’ is more relevant to the complicating requirements of
decision making in agricultural research (Contant and Bottomley, 1988). Shumway and McCracken
(1975), in their discussions on priority setting of agricultural research, were the first who used this
method in prioritizing plans the North California Agricultural Research Station. Similarly, Franzel
(1996), International Potato Center (CIP) and CGIAR have also used some combined methods.
Collion and Gregory (1993) combined the scoring model with the cost-benefit analysis for CIP
resource allocation. In addition a combination of the relevance (rule of thumb) method and the
scoring models were applied for CGIAR by McCalla and Ryan (1992).
However, Thomas L. Saaty suggests a method called ‘Analytical Hierarchy PROCESS’ (AHP) with
no such deficiencies of the scoring method while presenting all the advantages of participation,
transparency, and the standard procedure as well. Once the hierarchy is built, the decision makers
systematically evaluate its various elements, comparing them to one another in pairs. In making
the comparisons, the decision makers can use concrete data about the elements, or they can use
their judgments about the elements’ relative meaning and importance (Saaty, 2008). Currently
this technique is widely used in complicated management decision makings which, among others,
include: budgeting public administrations; transportation planning (Ferrari, 2003); planning
for energy resource allocation (Ramanathan and Ganesh, 1995); urban planning (Rose and
Anandalingam, 1996); setting priority for energy and environmental research projects (Kagazyo
et al., 1997); prioritization of electricity industries (Kaban, 1997); design of renewable energy
systems (Chedid et al., 1998); identification of favorable fuels in transportation industries (Poh
and Ang, 1999); evaluating machine tool alternatives (Ayag andOzdemir, 2006); and technology
assessment (Herkert et al., 1996). In recent years, the application of AHP method has also been
common in decisions related to the agricultural research management. Alphonce (1997) suggested
the AHP approach to agricultural research. Anders and Mueller (1995) also used this technique
to design long-term field experiments in International Crop Research Institute for Semi-Arid
32
EFITA conference ’09
Tropical (ICRISAT) and to identify public preferences for land preservation (Duke and Aull-hyde,
2002). Some other researchers have also applied the AHP method to selecting either an optimum
combination of research in Private sector (Libei-ator, 1989; Lockell et al., 1986; and Manahan,
1989) or a combination basket of agricultural research in Public sector ((ISNAR, 1998).
Methodology
This article illustrates how to use Analytical Hierarchy Process (AHP) as a well-known MultiAttribute Decision Making (MADM) method, and also how to combine qualitative and quantitative
indexes for priority setting of esearch programs in Sugar Beet Research Institute.
For this purpose, basics of priority setting for agricultural studies were reviewed; and followed by
a comparative study, an initial list of indexes and subindexes was specified for decision making,
which has subsequently been finalized by holding a professional poll. Then, through a decision
subject modeling, research programs were determined within the AHP model, representing a decision
hierarchy tree. Required data were gathered through a paired comparison questionnaire formulated
by the concerned experts and researchers. In different stages of estimation, ‘Expert Choice’ software
was applied and, eventually priority setting results were determined.
Implementation of analytical hierarchy process
Determining indexes and criteria of assessment
To identify and define indexes and criteria, one may take a number of different ways the most
significant of which includes conducting a comparative study and holding professional workshops
with experts and associated professionals. Braunschweig (2000) has used eleven indexes and
subindexes to set biotechnological research priority in Chile: In another research conducted by
ISNAR institute for Agricultural Research Institute of Kenya, the following indexes and criteria
were selected (ISNAR, 1998): Efficiency; Equity; Foreign exchange gains; Food self-sufficiency;
and Sustainability.
Therefore, the present study provided a preliminary list of indexes which was then finalized through
holding a poll session with elites and key experts in sugar beet researches. Accordingly, decision
hierarchy tree was drawn as presented in Figure 1.
Decision hierarchy tree
According to AHP approach, every decision making subjects can be explained within a hierarchical
structure known as decision hierarchy tree, in which the objective is at the first level and rival options
are at the last level while decision indexes are seen at the mid-level/s. Modeling decision making
is initially undertaken by applying AHP and drawing a decision hierarch tree.
As the concerned institute has been a ‘product research institute’ undertaking the projects focused
on a specific product, for the purpose of index determination, the indexes were made relevant and
some of them were eliminated.
Calculation stages of AHP method
Stage 1: paired comparisons
Following the formation of decision hierarchy tree, present components at each level are respectively
assessed from bottom-up levels relative to all the associated components at the higher levels.
Therefore, the assessments of decision options are carried out in terms of the last decision indexes
EFITA conference ’09
33
method to selecting either an optimum combination of research in Private sector (Libei-ator,
1989; Lockell, et al., 1986; and Manahan, 1989) or a combination basket of agricultural research
in Public sector ((ISNAR, 1998).
Figure 1: Decision Hierarchy Tree
Priority Setting for Research
Programs of Sugar Beet Seed
Economic
Development
Scientific
Development
Increased Food Security
Reduced Wastage
Improved Trade Balance
Maintained Employing
Development
Creating Value Added
Improved Productivity of
Production Resources
Achieved Knowledge and
Technology
Achieved New Resources,
Services, and Products
Research
Program 1
Research
Program 2
Research
Program 3
Number of Beneficiaries of
Study Results
Sustainable
Development
Inter-disciplinary Activity
Declined Pollution
Base Resource Conservation
Genetic Resource
Conservation
Research
Program 4
Declined Natural Disasters
Costs of Study
Duration of Study
Feasibility of
Study
Required Area, Laboratory,
and Equipments
Expertise Human Force
Duration of Study
Implementation
Participation of Beneficiaries
in Study
Conformity with Study
Orientations and Policies
Research
Program 5
Research
Program 6
Figure 1. Decision hierarchy tree.
which are also assessed in terms of their own hierarchy. In the AHP method, when the assessment
is based on quality, it is done in a paired comparison manner, where a square matrix is formed,
corresponding to the number of components which are placed in rows and columns. Then, these
options are compared with each other in a binary manner by decision makers and numerically
scored according to Sa’ati’s standardized table, and presented in the matrix columns.
Data matrix, A, is generally positive and reverse; and its components are indicated by aij. So,
considering the reversibility property of aij =1/aij, simply the comparisons by a number of n (n-1) /2
times are needed in a matrix of n.n. On the other hand, when the assessment is based on quantity,
the assessed components are measured by the same basis.
34
EFITA conference ’09
Data matrix, A, is generally positive and reverse; and its components are indicated by aij. So,
considering the reversibility property of aij =1/aij, simply the comparisons by a number of n (n-1)
/2 times
are needed
in aand
matrix
of n.n.inOn
rding to Sa’ati’s
standardized
table,
presented
thethe other hand, when the assessment is based on
quantity,
assessed
components
measuredtobySa’ati’s
the same
basis.
makerstheand
numerically
scoredareaccording
standardized
table, and presented in the
matrix
initsin
acomponents
group
making,
each
maker’s
obtained within
within the
thementioned
So,
acolumns.
groupdecision
decision
making,by
each
decision
maker’s viewpoint
viewpoint is
is obtained
and reverse;So,
and
are indicated
aij. decision
So,
mentioned
matrixes
then
combined
areverse;
group matrix.
For
the purpose
of
creatingbygroup
Data
isand
generally
positive
and
andFor
its the
components
arecreating
indicated
. So,
thematrix,
comparisons
by
a number
of nainto
(n-1)
of aij =1/aij, simply
matrixes
andA,then
combined
into
group
matrix.
purpose of
groupaijmatrixes,
as
matrixes,
shown
by assessment
Sa’ti and property
Aczel
(Forman
and Peniwati,the1998),
applyingbygeometric
mean
is
comparisons
a number
of
n best
(n-1)method.
considering
thethe
reversibility
of on
aPeniwati,
n.n. On the other
hand,asby
when
is based
ij =1/aij, simply
shown
Sa’ti
and
Aczel
(Forman
and
1998),
applying
geometric
mean
is
the
best
method.
/2
times
are needed in a matrix of n.n. On the other hand, when the assessment is based on
measured
bythe
the
same
basis.
rically
scored
according
to Sa’ati’s
table,
and
presented
in theconduction’ indexes were determined in
In
this
research,
thestandardized
‘cost
of study’
and
‘duration
of study
quantity,
the assessed
components
are‘duration
measuredofby
the same
basis. indexes were determined
In
this
research,
the ‘cost
of study’ within
and
study
conduction’
each decision maker’s
viewpoint
is obtained
thecomparisons
a
quantitative
manner
and
so,
no
paired
been
for them.
indexes are
in aSo,
quantitative
manner
and of
so,
no paired
comparisons
havehave
been
done done
for them.
These These
indexes
in aFor
group
decision
making,
each
decision maker’s
viewpoint
is obtained
within the
ed
into
a
group
matrix.
the
purpose
creating
group
generally positive
and
reverse;
and
its
components
are
indicated
by
a
.
So,
of
negative
aspect;
it
means
that
their
lower
values
get
higher
satisfactions.
So,
their
determined
ij get higher satisfactions. So, their
of negative
aspect;
it geometric
means
thatmean
their
values
mentioned
matrixes
and
then combined
into
a group
matrix.
For the purpose of creating group
l (Forman property
andare
Peniwati,
1998),
applying
isalower
, simply
the
comparisons
by
number
of manner.
n (n-1)
versibility
of aijwere
=1/aijreversely
determined
values
were
reversely
standardized
in
a
linear
The
standardized
values
were
values
standardized
in
a
linear
manner.
The
standardized
values
were
resulted
matrixes, as shown by Sa’ti and Aczel (Forman and Peniwati, 1998), applying geometric mean is from
ed in a matrix
of
On the
thevalues
other hand,
when
the
is based
on
resulted
from
reversal
values
divided
by
their
sum
(1.04).
In addition,
weighted
then.n.
reversal
divided
by
theirassessment
sum
(1.04).
In addition,
weighted
scorescore
was was
calculated by
the
best
method.
sed‘duration
components
are measured
by the
same
basis.coefficient
nd
of
study
conduction’
indexes
were
determinedof the concerned index (0.1) by the standardized
calculated
by
multiplying
weight
multiplying
weight
coefficient
of
the
concerned
index
(0.1)
by
the
standardized
values.
In have
this research,
thefor‘cost
of These
study’ indexes
and ‘duration of study conduction’ indexes were determined
aired comparisons
done
them.
values.
decision
making,
each been
decision
maker’s
viewpoint
is obtained within the
in
a quantitative
manner
and so, no
paired
have been done for them. These indexes
hat
their
lower
values
get
higher
satisfactions.
So,
theircomparisons
es and then combined into a group matrix. For the purpose
of creating group
are
of
negative
aspect;
it
means
that
their
lower values get higher satisfactions. So, their
Stage
2:
extracting
weight
coefficients
of
matrixes
ndardized
in
a
linear
manner.
The
standardized
values
were
Stage
2.
Extracting
Weight
Coefficients
of
Matrixes
n by Sa’ti and Aczel (Forman and Peniwati, 1998), applying geometric mean is
determined
values
were
reversely
standardized
in
a linear manner.
The
standardized
values
were
ided by theirInsum
(1.04).
Infirstly,
addition,
weighted
score
was
In this
stage,
firstly,
comparison
matrixes
normalized.
There
aremany
many
methods
forthis
this
purpose,
this
stage,
comparison
matrixes
areare
normalized.
There
are
methods
for
resulted
from
reversal
values
by their
sumdimensionless’,
(1.04).
addition,
weighted
score
efficient of the
concerned
index
(0.1)
bybythe
standardized
purpose,
asthe‘dimensionless
bydivided
Euclidean
norm’,
‘fuzzy In
dimensionless’,
and
‘linearwas
such assuch
‘dimensionless
Euclidean
norm’,
‘fuzzy
and ‘linear
dimensionless’,
the
he ‘cost of study’
and ‘duration
oflast
study
conduction’
indexes
were
calculated
bythe
multiplying
of the
concerned index (0.1) by the standardized
dimensionless’,
one
of weight
which
iscoefficient
used
in AHP
asdetermined
follows:
last
one
of
which
is
used
in
AHP
as
follows:
manner and so, novalues.
paired comparisons have been done for them. These indexes
a ij′
a ij lower values get higher satisfactions.
spect;
means that their
ents
of it
Matrixes
OR
rij So,
= n their,
j = 1, 2 ,..., m
rij = n
,
j = 1,..., m
were
reversely
standardized
in
a
linear
manner.
The
standardized
values
Stage 2.There
Extracting
Weight
Coefficients
atrixes are normalized.
are many
methods
for thisof Matrixes were
a ij′
a ij
∑
∑
valuesnorm’,
divided
by their
sum (1.04).
Inand
addition,
score
In thisi‘fuzzy
firstly,
comparison
matrixes
are normalized.
yreversal
Euclidean
‘linearweighted
i =1 wasThere are many methods for this
= 1stage, dimensionless’,
tiplying
coefficient
of the
index by
(0.1)Euclidean
by the standardized
purpose,
as concerned
‘dimensionless
norm’, ‘fuzzy dimensionless’, and ‘linear
is
used inweight
AHP as
follows:such
or
matrix
by which
coefficients, Wj , can be extracted.
Here,
rij is a normalized
dimensionless’,
the last
one ofcomponent,
which is used
in AHPweight
as follows:
a ij′
purpose,
there
are
a
few
methods
inclding
Anthropy,
Linmap, Lowest Weighted
OR Forrijthis
= n
,
j = 1, 2 ,..., m
a ij′
a ij
OR
Squares,
and
Specific
Vector
which
might
be
applied
(Hwang,
et
al.,
1995).
ng Weight Coefficients
of
Matrixes
r
=
,
j = 1, 2 ,..., m
′
rij ∑
= a ijn
,
j = 1,..., m
ij
n
tly, comparison matrixes
area normalized. There are many methods for this
i =1
′
a
∑
ij
∑
ij
i =1
Stageby3: Calculation
of Consistency
Rate
‘dimensionless
Euclidean
norm’,
‘fuzzy dimensionless’,
and ‘linear
i = 1 a normalized
is
component,
by which
weight
coefficients, wj, can be extracted. For
ponent, by whichHere,
weightrijcoefficients,
Wj , matrix
can be extracted.
Prior
to
analyzing
data,
consistency
of
comparisons
should
be
ensured,
since
theWeighted
factors were
e last one of which
is
used
in
AHP
as
follows:
this
purpose,
there
are
a
few
methods
inclding
Anthropy,
Linmap,
Lowest
Squares, and
matrix component,
by which weight coefficients,
Wj , can
be extracted.
Here,
rij is a normalized
methods inclding
Anthropy,
Linmap,
Lowest
Weighted
compared
by decision
makers
in
a
paired
series
and
they
are
likely
to
be
inconsistent
in general.
′
a
Specific
Vector
which
might
be
applied
(Hwang
et
al.,
1995).
ij
For this consistency
purpose,
are
a, be
fewpossible
inclding
Anthropy, were
Linmap,
Weighted
mightj be
(Hwang,
et
al., 1995).
OR
Calculating
rate nwould
when
the comparisons
done Lowest
on the basis
of
rthere
jmethods
= 1, 2 ,...,
m
= 1applied
,..., m
ij =
Squares,
and Specific Vectora ′which might be applied (Hwang, et al., 1995).
Saaty’s
scope.
∑
ij
Rate
iconsistency
=1
Stage 3: rate
calculation
ofby
rate rationale of specific vectors (Hwang, 1995).
Consistency
is
meared
a
mathematical
of Consistency
of comparisons Stage
should3:
beCalculation
ensured, since
the
factors Rate
were
Prior
to
analyzing
data,
consistency
of comparisons
should be ensured, since the factors were
malized matrix component,
by whichdata,
weight
coefficients,
, can be extracted.
Prioraretolikely
analyzing
consistency
of Wj
comparisons
should
the factors
ired series and
they
be inconsistent
i,j,k were
=
Mathematically,
iftocomponents
haveinageneral.
full
consistency,
we
will be
thenensured,
have: asince
ij = a kj × a ik
compared
by
decision
makers
in
a
paired
series
and
they
are likely
to
be inconsistent
in general.
arewhen
a few
methods
inclding
Anthropy,
Linmap,
Lowest
Weighted
compared
by
decision
makers
in
a
paired
series
and
they
are
likely
to
be
inconsistent
in
general.
e there
possible
the
comparisons
were
done
on
the
basis
of
1,2,…,n
. So, ifconsistency
all the components
of the
matrix Awhen
show the
a full
consistency,
we done
will have:
Calculating
rate
would
be
possible
comparisons
were
on
the
ific Vector whichCalculating
might be applied
(Hwang,
al., 1995).
consistency
rateetwould
be possible when the comparisons were done on the basis ofbasis of
CI
w
i
Saaty’s
scope.
scope.
.The
consistency rate is: CR =
, in which RI is the random index, suggested by
aij =Saaty’s
ematical rationale
ofj specific vectors (Hwang, 1995). RI
w
ion of Consistency
Rate
Consistency
rate
is meared
a mathematical
specific
vectors
(Hwang, 1995).
Consistency
rate
is meared
by a by
mathematical
rationalerationale
of specificofvectors
(Hwang,
1995).
Saaty, of
in proportion
of should
dimensions.
consistency
comparisons
factors were
i,j,ksince
adata,
full consistency,
we
will then have:
amatrix
a kj ×ensured,
aik have
ij = be
Mathematically,
ifthe
components
a =full the
consistency,
we will then have: aij = akj× aik with i,j,k =
Mathematically,
ifthey
components
full consistency,
we will then have: aij = a kj × aik i,j,k =
sion
makers
in aA
paired
are likely
to beahave:
inconsistent
in general.
of the
matrix
showseries
a So,
fulland
we have
will
1,2,…,n.
ifconsistency,
all thecomparisons
components
of the
matrix
A basis
showofa full consistency,
we will have:
tency
be possible
done
on the
. So,when
if allthethe
componentswere
of the
matrix
A show
a full consistency, we will have:
CI rate would1,2,…,n
R=
, in which RI is the random index, suggested by
wi .The consistency rate is: CR = CI , in which RI is the random index, suggested by
RI
aij =
sensions.
meared by a mathematical
rationale of specific vectors (Hwang,
RI 5 1995).
wj
Saaty,
in consistency,
proportion ofwe
thewill
matrix
components have
a full
thendimensions.
have: aij = a kj × aik i,j,k =
The consistency rate is:
all the components of the matrix A show a full consistency, we will have:
CI
istency rate is: CR =
, in which RI is the random index, suggested by
RI
5
5
n of the matrix dimensions.
in which RI is the random index, suggested by Saaty, in proportion of the matrix dimensions.
To apply a group AHP to a combination of individual matrixes, the geometric mean is used and as
a result, the consistency rate of comparisons will be greatly reduced. To select the best options or
5 all the w s of rival options are multiplied by the w s of the coresponding decision
prioritize them,
i
i
indexes, resulting in the weighted mean of each option. Finally an option with the highest weighted
mean is set as the best option and other options are placed at next priorities. Obviously, as the
EFITA conference ’09
35
study programs and projects were assessed by fifteen indexes, the same number of categories wi
were produced, as shown by Table 1; accordingly, the achieved priority of each program is also
presented at the bottom of the table. As indicated in Table 1, the sugar beet research priorities
respectively include indexes of sustainable developmeent, scientific development, feasibility of
study, and economic development. In addition, the highest priorities arefound for certain subindexes
including genetic resource conservation (0.14), base resource conservation (0.136), knowledge and
technology achievement (0.131), and achieving new products and services (0.113). Accordingly,
research program priorities of the Sugar Beet Research Institute were respectively specified as
follows: Rhizomonia and Curly-Top disease (0.276); environmental tensions of the crop (salinity
and draught) (0.16); quality and quantity of the crop (o.166); examination of Rhizomania resistance
under glass-coverage and infectec area conditions (0.137); examination of the main agronomicphysiological properties (o.135); and finally, qualitative and quantitative study of tha crop within
both furrow and micro irrigation systems (0.118).
Discussion and conclusion
An entirely comprehensive coverage of issues, problems and demands in an agricultural research
system would be generally impossible in effect, due to either a broad multiple range of the
beneficiaries concerned in such studies, resulting in their varied demands and expectations, or a
number of considerable limitations such as financial and time constraints. Therefore, it requires
an optimal use of the existing potentials, resources and capacities through research planning and
priority setting so as to maximize research sources to a certain level of costs and/or minimize costs
to a certain level of sources.
However, in effect, the priority setting and approval of research programs and projects within
agricultural research system are far from efficiency and impact by many reasons including vague
research policies and priorities, numerous involved authorities and institutions, unrealistic fund
allocations of programs and projects, and governing bureaucratic procedures in the priority setting
process. Each one of these institutions and authorities tends to study the need for research in its
own viewpoint which is not only inconsistent but also varied and, in some cases, conflicting with
another, resulting in certain negative consequences.
Indeed, applying appropriate methods of priority setting seems to be a prerequisite to make efficient
the priority setting process of research programs and projects. For this purpose, there might be used
different methods; and among others, multiple index decision-making methods are now widely
used in various contexts, resting on their high capabilities in modeling real issues, simplicity and
understandability for users as well as taking account of both quantitative and qualitative conditions
and variables of an issue at same time. So, this article presents how to apply the Analytical Hierarchy
Process (AHP) to the agricultural research projects. Produced results are of great importance in
illustrating group decisions more explicitly and make contingency in the views of decision-making
group; thus, conflicts and controversies in dominant views are avoided and the adopted decisions
are more likely to be enforced.
Of course, it should be noticed that like any other methods in decision-making, these techniques
tend to simply turn data into information and provide decision-maker with them; and so, it is up to
the decision-maker to make optimum decision under organizational situations and circumstances
according to the produced results, and avoid to absolutely adopting the results. Therefore, it is
frequently suggested that workshops involving decision-makers are set up in order to analyze the
produced results and make a final decision. Moreover, since the conditions and factors effective on
agricultural research priority setting are growing and complicating under the influence of increasing
36
EFITA conference ’09
EFITA conference ’09
37
Economic
Increased food security (0.381)
development Reduced wastage (0.219)
Improved trade balance (0.127)
(0.194)
Improved productivity of production resources
(0.272)
Scientific
Knowledge and technology achievement (0.433)
development Achieved new resources, services and products
(0.373)
(0.304)
Number of Research Beneficiaries (0.195)
Sustainable
Declined pollution (0.197)
development Base resource conservation (0.395)
Genetic resource conservation (0.408)
(0.343)
Feasibility of Costs of study (0.1)
study (0.159) Duration of study (0.1)
Required area, laboratory, and equipments (0.249)
Participation of beneficiaries in study (0.142)
Conformity with research orientations and policies
(0.309)
Weighted mesn scores
Final priority
Decision indexes and weight coefficients
Table 1. Research program priority results.
Assessment and prioritization of sugar beet
research programs
0.01
0.02
0.03
0.02
0.01
0.02
0.03
0.02
0.02
0.01
0
0
0.02
0.01
0.01
0.074
0.042
0.025
0.053
0.131
0.113
0.059
0.068
0.136
0.14
0.032
0.016
0.040
0.023
0.049
0.095
0.175
0.255
0.061
10.034
00.867
0.154
0.113
0.084
0.135
5
0.276
1
0.163
0.069
0.095
0.124
0.064
0.125
2
0.357
0.322
0.138
0.318
0.108
0.079
0.117
0.330
0.329
0.311
0.358
0.384
0.259
0.421
0.256
1
0.169
2
0.173
0.099
0.167
0.207
0.284
0.079
0.205
0.154
0.145
0.158
0.170
0.149
0.127
0.169
0.178
3
0.166
3
0.172
0.135
0.123
0.196
20.082
00.868
0.293
0.119
0.152
0.107
0.171
0.163
0.301
0.140
0.197
4
0.118
6
0.095
0.110
0.217
0.063
0.147
0.476
0.122
0.103
0.088
0.106
0.078
0.069
0.069
0.081
0.124
5
0.137
4
0.108
0.159
0.100
0.155
0.148
0.191
0.109
0.181
0.202
0.155
0.127
0.139
0.120
0.126
0.120
6
Total Consistency Weighted score of rival options in terms of
weight rate
related indexes
developments and changes, and as little simplifications in modeling decisions should be made to
allow their improvement, application of a phased AHP is recommended. On the other hand, by
using other multiple index decision-making methods including ‘TOPSIS’ and ‘ELECTRE’, we
can provide different scenarios of priorities and achieve considerable results for decision-makers
by comparing them. In this context, analyzing result signification can help explaining strengths
and weaknesses of each method and presenting a practice to adopt the most appropriate method in
terms of the existing conditions. This is suggested as one of the research grounds
References
Alphonce, C. B. (1997), Application of the analytic hierarchy process in agriculture of developing countries.
Agricultural System, 53(1): 97-112.
Anders, M. M. and Mueller, R. A. E. (1995), Managing communication and research task perceptions in
interdisciplinary crops research. Quarterly Journal of International Agriculture. (34): 53-69.
Ayag Z., Ozdemir R. G. (2006), A fuzzy AHP approach to evaluating machine tool alternatives. Journal of
Intell Manuf. (17): 179-190.
Bose, R. K. G. and Anandalingam (1996), Sustainable urban energy-environment management with multiple
objectives. Energy. 21 (4): 305-318.
Braunschweig, T. (2000), Priority Setting in Agricultural Biotechnology Research. ISNAR.
Chedid, R., Akiki, H., and Rahman, S. (1998), Decision support technique for the design of hybrid solar-wind
power systems. IEEE Trans. Energy Convers. 13 (1): 76-83.
Duke J. M. and Aull-hyde R. (2002), Identifying public preferences for land preservation using the analytic
hierarchy process. Ecological Economics. (42): 131-145.
Ferrari, P. (2003), A method for choosing from among alternative transportation projects. European Journal
of Operational Research. (150): 194-203.
Forman, Ernest and Peniwati, Kirti (1998), Aggregating individual judgment and priorities with the analytic
hierarchy process. European Journal of Operation Research. (108): 156-169.
Herkert, J. R., Farrell, A., and Winebrake, J. J. (1996), Technology choice for sustainable development. IEEE
Technol. Soc. Mag.. 15 (2): 12-20.
Hwang, C. L. and Yoon, K. P. (1995), Multiple Attribute Decision Making: An Introduction. London: Sage
Publications.
ISNAR (1998), Agricultural Research Priority Setting: Information Investments for Improved Use of Resources.
Edited by Bradford Mills.
Kablan, M.(1997), Prioritization of decentralized electricity options available for rural areas in Jordan. Energy
Convers. Manage.. 38 (14): 1515-1521.
Kagazyo, T., Kaneko, K., Akai, M., and Hijikata, K. (1997), Methodology and evaluation of priorities for
energy and environmental research projects. Energy. 22 (2/3): 121-129.
Liberatore, M. J. (1989), A decision approach for R&D project selection. In: The Analytic Hierarchy Process:
Applications and Studies. Edited by B. L.
Lockett, G., Hetherington, B., Yallup, P., Stratford, M., and Cox, B. (1986), Modeling a research portfolio
using AHP: a group decision process. R&D Management. 16(2): 151-60.
Manahan, M. P. (1989), Technology acquisition and research prioritization. International Journal of Technology
Management. 4(1): 9-19.
Pardey, P. G. and Roseboom, J. (1998), Development of national agricultural research systems in an international
quantitative perspective. In: Technology Policy for Sustainable Agricultural Growth, IFPRI Policy Briefs.
No. 7. Washington, D.C.
38
EFITA conference ’09
Poh, K. L. and Ang, B. W. (1999), Transportation fuels and policy for Singapore: an AHP planning approach.
Comput. Ind. Eng.. 37: 507-525.
Ramanathan, R. and Ganesh, L. S. (1995), Energy resource allocation incorporating qualitative and quantitative
criteria: an integrated model using goal programming and AHP. Socio-Econ. Plann. Sci.. 29 (3): 197-218.
Saaty, T.L. (2008-06), Relative measurement and its generalization in decision making: why pairwise
comparisons are central in mathematics for the measurement of intangible factors- the analytic hierarchy/
network process. RACSAM (Review of the Royal Spanish Academy of Sciences, Series A, Mathematics.
102 (2): 251-318. Available on: http://www.rac.es/ficheros/doc/00576.pdf. Retrieved on 22 December 2008.
EFITA conference ’09
39