1
Assessing farming eco-efficiency:
A Data Envelopment Analysis approach
Andrés J. Picazo-Tadeo, José A. Gómez-Limón and Ernest Reig-Martínez*
Abstract.- This paper assesses farming eco-efficiency using Data Envelopment Analysis techniques.
Eco-efficiency scores at both farm and environmental pressure-specific levels are
computed for a sample of Spanish farmers operating in the rain-fed agricultural system of
Campos County. The determinants of eco-efficiency are then studied using truncated regression
and bootstrapping techniques. Our results reveal that farmers are quite eco-inefficient,
with very few differences emerging among specific environmental pressures. Moreover, ecoinefficiency
is closely related to technical inefficiencies in the management of inputs. Regarding
the determinants of eco-efficiency, farmers benefiting from agri-environmental programs
as well as those with university education are found to be more eco-efficient. Concerning the
policy implications of these results, public expenditure in agricultural extension and farmer
training could be of some help to promote integration between farming and the environment.
Furthermore, CAP agri-environmental programs are an effective policy to improve ecoefficiency,
although some doubts arise regarding their cost-benefit balance.
Keywords.- Farming; economic-ecological efficiency; environmental pressures; Data
Envelopment Analysis; bootstrapping.
JEL Classifications.- C61, D21, Q56.
1. INTRODUCTION
The vital role of agriculture in providing food and fibre to a rising human population
has made of this productive activity a privileged field for sustainability analysis. But
more than twenty years after the publication of the Brundtland Report, which famously
defined sustainable development as ‘development that meets the needs of
the present without compromising the ability of future generations to meet their own
needs’ (WCED, 1987: 43), sustainability in general, and agricultural sustainability in particular,
remains an elusive concept. This explains why an impressive amount of research
has been undertaken to overcome conceptual vagueness by defining an appropriate
scale of reference and by developing composite indicators covering socio-
*
Andrés J. Picazo-Tadeo and Ernest Reig-Martínez belong to the Department of Applied Economics
II. University of Valencia. Avda. dels Tarongers s/n, 46022 Valencia (Spain). José A.
Gómez-Limón works at the Department of Agricultural Economics. Instituto Andaluz de Investigación
y Formación Agraria y Pesquera. PO Box 3092. E-14080 Córdoba (Spain). The research
was co-financed by the Spanish Ministry of Science and Technology (research projects
AGL2006-05587-C04 and AGL2009-12553-C02) and the Regional Government of Castilla y León
(research project VA036A08). Andrés J. Picazo-Tadeo also acknowledges the financial aid
received from the Spanish Ministry of Science and Technology (project ECO2008-05908-C02-
02) and the Generalitat Valenciana (program PROMETEO 2009/098).
2
economic and environmental issues (Van der Werf and Petit, 2002, Böhringer and
Jochem, 2007, Van Cauwenberg et al., 2007, Bell and Morse, 2008). Experts in this field
have argued that developing sustainability indicators ‘pulls the discussion of sustainability
away from abstract formulations and encourages explicit discussion of the operational
meaning of the term’ (Rigby et al., 2000: 5).
Assessing agricultural sustainability at farm level, as we do in this paper, is a particularly
difficult task, as no consensus exists concerning the relevant environmental
variables, while at least some standardisation has been achieved for assessments undertaken
at national or macro-level (OECD, 2001 or EEA, 2005). A workable approach
to sustainability at farm level consists in evaluating whether individual farmers are
making efficient use of natural resources in order to achieve their economic objectives.
Economic-ecological efficiency, commonly known as eco-efficiency, emerged
in the 1990s as an operational concept to allow for a practical approach to sustainability
(Schaltegger and Sturm, 1990, Schaltegger, 1996). It was adopted and popularised
by the World Business Council for Sustainable Development (Schimedheny with
the BCSD, 1992, WBCSD, 2000) as a way to encourage companies to become simultaneously
more competitive and more environmentally responsible. The topic was
also addressed by the OECD (1998), which defined eco-efficiency as ‘the efficiency
with which ecological resources are used to meet human needs’. Accordingly, ecoefficiency
can be measured by using ratios that relate the economic value of products
and services that account for the output of a firm, an industrial branch or a territory,
to the sum of environmental pressures or impacts involved in the production
process. Eco-efficiency improves when environmental impacts decrease as the value
of economic outputs is maintained or increased.
Eco-efficiency serves two broad goals. At macro-level it is a reminder that GDP
growth should be de-linked as much as possible from its potential negative environmental
impacts, as human societies aspire to the satisfaction of rising consumption
levels and the simultaneous attainment of reasonable environmental quality. At micro-level
it means creating more value with less environmental impact. The literature
suggests alternative measures for the eco-efficiency ratio depending on the scale of
analysis, the adoption of a short or long-term perspective and the broadness of scope
in the definition of both economic value and environmental impact. The simplest indicators
measure economic output per unit of waste, but others include not only value
added, but also jobs and social welfare indicators in the numerator. Concerning the
denominator, pressure indicators, such as CO2 emissions and natural resource consumption,
and environmental impact indicators, including greenhouse effects or
ozone depletion, have been considered.
3
An improvement in the eco-efficiency coefficient does not necessarily guarantee
sustainability, because what this coefficient measures is only the relative level of environmental
pressure in relation to the volume of economic activity. And what really
counts when dealing with sustainability issues is absolute rather than relative environmental
pressure, which can still exceed the carrying capacity of the ecosystem. It has
been emphasized that eco-efficiency improvements at micro-level do not guarantee
that environmental quality goals at macro-level can be reached (Huppes and Ishikawa,
2005). A global rise in consumption, stemming from economic growth over time
can nullify the eco-efficiency gains obtained per unit of consumption. This is particularly
worrying if the consumption pattern shifts in the direction of stronger environmental
impacts, as may happen when the share of animal products in the diet increases,
or when average consumer spending on travel and energy-consuming domestic
appliances continues to climb.
In spite of the abovementioned criticisms, measuring eco-efficiency remains important
at least for two reasons (Kuosmanen and Kortelainen, 2005). First, as an improvement
of eco-efficiency is often the most cost-efficient way of reducing environmental
pressures, and second, because policies that target improvements in ecoefficiency
are easier for policymakers to implement than more drastic measures
aimed at restricting the level of economic activity. In fact, a win-win outcome from
policies promoting eco-efficiency can be frequently expected. It obeys to companies
often not operating at the frontier of economic efficiency, which creates a chance of
making net cost savings in addition to reducing their environmental impacts (Ekins,
2005).
The objective of this paper is to assess the eco-efficiency of farmers operating in
the rain-fed agricultural systems of the Castilla y León region, in Spain. In doing so, we
adopt a micro-level environmental-productivity ratio, following the classification in
Huppes and Ishikawa (2005). More specifically, eco-efficiency is defined as the ratio
between value added and a composite indicator of environmental pressures. Value
added is computed at farm level and includes the so-called coupled payments of
the post-2003 Common Agricultural Policy (CAP) of the European Union, and agrienvironmental
payments. The composite indicator of environmental pressures, also
quantified at farm level, comprises five environmental variables: specialization, nitrogen
and phosphorus balances, risk of pesticides and energy balance. Eco-efficiency
scores at farm level and pressure specific eco-efficiency scores, also at farm level, are
computed using Data Envelopment Analysis (DEA) techniques. The work proposed in
this way is interesting both from an academic point of view and from a policy-making
perspective, considering that the results obtained may help to re-design agrienvironmental
policy measures in order to reach a better benefit-cost balance.
4
Following this introduction, the next section briefly reviews the DEA literature in the
field of eco-efficiency measurement. Section three expounds the main insights of the
methodology. Section four describes the agricultural system analysed and comments
on the data. Section five discusses the main findings and puts them into context, and,
finally, section six concludes.
2. A BRIEF COMMENT ON DEA IN THE CONTEXT OF ECO-EFFICIENCY MEASUREMENT
Data Envelopment Analysis is a non-parametric methodology, pioneered by Charnes
et al. (1978), aimed at evaluating the relative efficiencies of comparable decisionmaking
units (hereafter DMUs) by means of a variety of mathematical programming
models. One recognised advantage of DEA is that no prior assumptions concerning
the specific functional relationship linking inputs and outputs are imposed. Instead, a
piecewise linear frontier is constructed based on empirical observations on inputs and
outputs of a sample of DMUs. The technological frontier represents best practices,
while the distance to it from each DMU in the sample is used to compute a measure
of its relative performance (see Cooper et al. 2007, and Cook and Seiford, 2009 for an
appraisal of the theoretical foundations and developments in DEA).
Conventional DEA analysis allows the researcher to assess the performance of individual
DMUs taking only into account observed quantities of marketable inputs and
outputs. However, as the field of DEA applications has progressively grown, a distinctive
research stream has focused on employing this technique to address the environmental
consequences of production processes. Researchers have been compelled
to handle not only conventional outputs and inputs in their models, but also
bad or environmentally undesirable outputs, i.e., wastes and polluting effluents obtained
as by-products of commercial outputs, and inputs. A number of authors have
surveyed the main approaches adopted in the literature (Tyteca, 1996, Allen, 1999,
Scheel, 2001, and Zhou et al., 2008).
Korhonen and Luptacik (2004) expound the two different approaches used to incorporate
eco-efficiency in DEA models. The first requires performing a first step in
which DEA is used to compute separate evaluations of technical and ecological efficiency,
and then these efficiency figures are used as output variables in a new DEA
model. The second way consists of building up a ratio which simultaneously takes into
account both desirable and undesirable outputs, and leads to a wide variety of alternative
models, depending on how undesirable outputs are treated.
A variety of empirical applications have used DEA to assess eco-efficiency. On the
one hand, De Koeijer et al. (2002) calculate environmental efficiency scores with an
5
input-oriented DEA model using observed environmental inputs instead of conventional
inputs, then compute profit-efficiency and, finally, use the results from both calculations
to implement a DEA model of sustainable efficiency for a sample of sugar
beet farms in The Netherlands. Korhonen and Luptacik (2004) describe several alternative
models to assess eco-efficiency and test their ability to provide similar efficiency
scores for a sample of European power plants. Kortelainen and Kuosmanen
(2005) analyse the eco-efficiency of road transportation in three towns of Finland, after
aggregating several air emissions into four types of environmental pressures. Also
Zhang et al. (2008) use the linear programming transformation of a conventional multiplier
CCR model (Charnes et al., 1978), and the linear transformation of a ratio
model, where undesirable outputs are treated as inputs, to evaluate the ecoefficiency
of thirty provincial industrial systems in China.
3. METHODOLOGICAL ISSUES
In this paper we define eco-efficiency as a ratio between economic value added
and environmental pressures. Let us therefore assume that we observe the economic
value added, denoted by variable v, generated in the production processes by a set
of k = 1,…,K farms. In addition, the production process generates a set of n = 1,…,N
damaging environmental pressures, also observed at farm level, which are denoted
by variables pn. The pressure generating technology set (PGT) representing all feasible
combinations of value added v and environmental pressures p = (p1,…,pn) can be
defined as:
  N
PGT v p R v p1
, value added can be generated with pressures

    (1)
Following Kuosmanen and Kortelainen (2005), eco-efficiency of farm k can be
formally defined as:
 
 
k
k k
k
v
Eco-efficiency Eco-eff
P p
 , (2)
P being the pressure function that aggregates the n environmental pressures into a
single environmental pressure score.
While value added can be either directly observed or indirectly computed using
data on prices and quantities of outputs and intermediate inputs, constructing the
composite environmental pressure score is trickier. Kuosmanen and Kortelainen (2005)
pointed out that a reasonable approach to computing this score is to take a
weighted average of the particular pressures exercised by farm k on the environment,
that is:
6
 
N
k n nkn
P p w p1
  , (3)
where wn is the weight with which pressure n enters into the computation of the environmental
pressure score.
Building a composite indicator requires, therefore, the adoption of a weighting
scheme that should represent the relative importance of the different environmental
pressures. As no self-evident pattern of weights is available, any choice unavoidably
requires subjective valuation a priori. In order to avoid the bias stemming from a subjective
choice of common weights, in this paper we have decided in favour of using
DEA as our preferred aggregation method. Instead of a common scheme of weights,
this technique allows weights to be determined at farm level. In particular, the set of
weights for farm k is chosen so that it maximizes the relative eco-efficiency score of
this farm when it is compared to the others farms in the sample.
Formally, using DEA the eco-efficiency score of each farm k’ belonging to our
sample of k = 1,…,K farms is computed from the following programming problem:
nk
k
w k N
nk nkn
k
N
nk nkn
nk
v
Maximize Eco-eff
w p
v
subject to: k K i
w p
w n N ii
'
'
'
' '1
'1
'
1 1,..., ( )
0 1,..., ( )



 
 


, (4)
wnk’ being the weight with which pressure n enters into de computation of the composite
environmental pressure score of farm k’.
The problem (4) has an equivalent dual formulation, which can be written as:
k kz k k
K
k k kk
K
k nk k nkk
k
Minimize Eco-eff
subject to: v z v i
p z p n N ii
z k K iii
' , ' '
' 1
' ' 1
( )
1,..., ( )
0 1,..., ( )



 

  
 


, (5)
zk being, in this case, a set of intensity variables representing the weighting of each
observed farm k in the composition of the eco-efficient frontier. In other words, this
new set of weights allows us to compute a virtual eco-efficient point of reference for
farm k’.
The solution to this problem for farm k’, namely the parameter *k’, measures the
potential proportional reduction of all environmental pressures that it could achieve
7
while maintaining its value added, so that the resulting combination of value added
and environmental pressures belongs to the pressure generating technology set. By
construction, this score of eco-efficiency is upper-bounded to one, the score that
represents best performance. Moreover, the lower the score computed, the lower
eco-efficiency.
Making a parallelism with conventional DEA literature, scores computed from expression
(5) would measure eco-efficiency in a Farrell-Debreu sense (Farrell, 1957), as
they are assessing equiproportional or radial reductions of environmental pressures
necessary to attain eco-efficiency. Nevertheless, once the maximum proportional
reduction of all environmental pressures has been attained, additional reductions
may still be feasible in some pressure directions, while maintaining the value added.
These pressure-specific potential reductions can be computed from the following optimizing
program (Ali and Seiford, 1993):
' '
' ' ', , 1
' ' 1
*
' ' ' 1
' '
subject to: ( )
1,..., ( )
, 0 1,..., ( )
0 1,..., ( )



 
 
   
 
 



v p
k knk
Nv p
k k nks s z n
Kv
k k k kk
Kp
k nk nk k nkk
v p
k nk
k
Maximize S s s
v s z v i
p s z p n N ii
s s n N iii
z k K iv
, (6)
sp and sv representing pressure excesses and value added shortfalls, respectively.
The objective of program (6) is to find a solution that maximizes the sum of pressure
excesses and valued added shortfalls for each farm while keeping their radial ecoefficiency
scores at the level calculated from expression (5).
Potential proportional reductions of environmental pressures in addition to pressure
excesses can be used to assess pressure-specific eco-efficiency by adapting the
methodology proposed by Torgersen et al. (1996) to assess input-specific technical
efficiency. Prior to computing pressure-specific scores of eco-efficiency, both aggregate
pressure potential reductions and their efficient levels must be calculated. The
aggregate reduction of pressure n needed to bring farm k’ into a Pareto-Koopmans
efficient status (Koopmans, 1951) is computed by adding up radial reductions and
pressure-specific excesses:
  p
nk k nk nkp p sreduction *
' ' ' '1    (7)
The first term on the right hand side of expression (7) measures proportional reduction
of pressure n, while the second term quantifies the slack in the direction of this
8
environmental pressure, i.e., pressure-specific excess. Likewise, the Pareto-Koopmans
efficient level of pressure n is computed by subtracting its potential aggregate reduction
from observed level, yielding:
 Pareto Koopmans efficient p p
nk nk k nk nk k nk nkp p p s p s* *
' ' ' ' ' ' ' '1
         
(8)
Finally, the pressure-specific measure of eco-efficiency for farm k’ and pressure n is
computed as the quotient between the eco-efficient level of that pressure and its
actually observed level:
Pareto Koopmans efficient p
nk nk
k
nk nk
p s
p p
*' '
nk' '
' '
Pressure-specific eco-efficiency

    (9)
By construction, scores of pressure-specific eco-efficiency are equal to or lower
than radial scores. Moreover, they are upper bounded to one. A score equal to one
for pressure n points to eco-efficiency and means that no reduction is feasible without
decreasing value added. Conversely, computed scores of less than one represent
eco-inefficiency, the lower the score the greater the eco-inefficiency.
Including information about slacks in the assessment of eco-efficiency, as we propose
in this paper, reveals the full potential for reducing pressures on the environment
while maintaining value added. When the number of dimensions is large relative to
the number of observations, slacks might be picking up an important part of total potential
environmental pressure reductions, and pressure-specific measures of ecoefficiency
provide a largely enhanced picture of performance. Furthermore, the importance
of slacks in explaining pressure-specific eco-efficiency can be assessed by
computing the weighting of potential pressure reductions due to slacks, on total pressure
potential reductions. Formalising for pressure n:
 
 
 
 
p
nk
p
k nk nk
p p s
p p p s
K Kradial Pareto-Koopmans efficient
nk nkk=1 k=1
n K KPareto-Koopmans efficient *
nk nkk=1 k=1
=
1

 
     
 
 
, (10)
radial
nk k nkp p*
  being the pressure n that would result from the radial contraction of all
environmental pressures of farm k towards its eco-efficient reference on the frontier.
4. DATA AND SAMPLE
4.1. Case study: Rain-fed agriculture in Castilla y León (Spain)
The empirical application of the methodology proposed has been implemented on a
representative sample of 171 farms belonging to the rain-fed agricultural system of
9
Campos County, in the province of Palencia, located in the central part of the Spanish
North Plateau (about 800 m.a.s.l.). Characterized by a continental climate, production
in the region is mainly based on annual extensive crops, particularly winter
cereals.
This county has a surface area of 304,483 hectares, of which 86% (261,505 hectares)
is considered as utilised agricultural area (UAA). 83% of this UAA (254,992 hectares) are
rain-fed lands, where the major crops are: barley (52%), wheat (26%), alfalfa (5%), sunflower
(4%), oats (4%) and pulses (3%). We chose this agricultural system for our case
study firstly for practical interest, bearing in mind that it can be treated as a representative
case of extensive (low-input-low-output) agriculture, where environmental functions
play a relevant role (Kallas et al., 2007). Second, homogeneity of farms and the
fact that data are readily available are convenient features for efficiency analysis.
Furthermore, some farmers are receiving payments from CAP agri-environmental programs,
which allows us to study the relationship between eco-efficiency and agrienvironmental
policies.
4.2. Variables used in the analysis
Value added variable
As pointed out above, the product variable for the eco-efficiency analysis proposed is
economic value added per hectare at farm level (vk). In order to calculate this variable
for each farm in the sample, the following formula has been used:
  

cosk k k k
k
k
Sales Coupled subsidies Agrienvironmental payments Intermediate ts
v
Land
, (11)
where Sales comprises all incomes obtained from the sale of agricultural products,
Coupled subsidies are the subsidies of the CAP received by producers based on their
crop-mix, Agri-environmental payments are the payments received for those farmers
signing agri-environmental contracts, and Intermediate costs are the costs due to the
following inputs: seeds, nitrogen and phosphorous fertilizers, pesticides and energy.
Each of these variables has been taken individually for every farm k, being measured
in constant euros of 2008. Thus, the numerator in (11) measures the absolute value
added, in euros, of farm k obtained in 2008. This value added is expressed per hectare.
i.e., absolute value added divided by farm size measured in hectares, variable
Land.
Regarding this point, the consideration of CAP financial support (coupled subsidies
and agri-environmental payments) in the estimation of the value added must be clarified.
In this sense, it must be pointed that the consideration of farms as decision-
10
making units means that efficiency analysis has to be performed taking into account
the private point of view of farmers. It seems obvious that an efficient farmer is not
only one that uses inputs rationally, but also the one that chooses an adequate cropmix,
with a view to maximizing profits. Thus, taking into account that profitability depends
both on sales revenue and the subsidies linked to sowing decisions (as is the
case with CAP coupled subsidies) or linked to voluntary management agreements (as
is the case with agri-environmental payments), it is reasonable that both sources of
income, sales and linked subsidies, had been considered to calculate the value
added of farms.
Environmental pressure variables
Taking into account the ecological features of the agricultural system under consideration
(Gómez-Limón and Sánchez-Fernández, 2010), five indicators have been
deemed relevant to measure environmental pressures from farming activities:
1. Specialisation. This indicator quantifies the tendency of the farm towards monoculture,
and is measured as the percentage of a farm’s surface covered by the most
important crop:
1
k
k k
k
Main crop
p Specialisation
Land
  , (12)
where Main cropk and Landk are, respectively, the surface devoted to the main
crop in farm k and its total surface cultivated, both measured in hectares. To understand
the meaning of this indicator from an environmental perspective, it is
worth pointing out that excessive productive specialisation (high value of Specialisation)
is negative, because this involves a loss of biodiversity (flora and fauna
species) associated to the existence of different crops and field margins between
crops.
2. Nitrogen balance. This indicator quantifies the physical difference between the
nitrogen contained in inputs (fertilizers) and absorbed in outputs (harvest), measured
in kg of N per hectare:
C
c c ckc
k k
k
Nitrogen input Nitrogen output Land
p Nitrogen balance
Land
1
2
( )

 
 , (13)
where Nitrogen inputc is the quantity of nitrogen used to fertilize the crop c and Nitrogen
outputc is the quantity of nitrogen extracted when crop c is harvested, both
measured in kg of N per hectare. Finally, Landck is the surface area that farm k devotes
to crop c, measured in hectares. Thus, this balance provides the quantity of
nitrogen per hectare that is released to the environment every year, involving wa-
11
ter pollution (eutrofization). This indicator quantifies the contribution of the farming
sector to non-point pollution.
3. Phosphorus balance. Similar to the previous indicator, phosphorus balance, measured
in kg of P per hectare, is calculated as follows:
C
c c ckc
k k
k
Phosphorus input Phosphorus output Land
p Phosphorus balance
Land
1
3
( )

 
 (14)
where Phosphorus inputc is the quantity of phosphorus incorporated into the soil
through the fertilizers used for crop c and Phosphorus outputc is the amount of
phosphorus removed when crop c is harvested, both measured in kg of P per hectare.
The balance for phosphorus enables us to calculate the amount of this pollutant
element released into the ecosystem.
4. Pesticide risk. This indicator provides information about the overall toxicity released
into the environment through the pesticides used for the agricultural production.
This toxicity has been estimated by means of the potential lethality of live organisms
by the active matters included in these agrochemicals, measured in kg of rat
per hectare. The formula used with this purpose is:
 
 
 
C M mc
ckc m
m
k k
k
Quantity comercial product
Land
Lethal dose50
p Pesticide risk
Land
1 1
4
1.000
, (15)
where Quantity commercial productmc is the quantity of the product m applied to
crop c (in kg of m per hectare) and Lethal dose50m is the lethal dose 50% of the
commercial product m (in mg of m per kg of rat)1. Thus, as the value of this indicator
rises, the biocide effects of the farming sector also increase.
5. Energy balance. This indicator is the ratio between the energy consumed to generate
the agricultural inputs and the energy fixed by crops and exported by the
output harvested:
C
c ckc
k k C
c ckc
Energy inputs Land
p Energy ratio
Energy outputs Land
1
5
1
( )
( )


 


, (16)
where Energy inputsc is the amount of energy virtually included in the inputs used
for crop c and Energy outputsc is the energy included in the production of crop c,
both these variables being calculated in kcal per hectare. Thus, the higher the
value of this indicator, the less efficient farm k is from an energy perspective (more
1 The lethal dose 50% is the amount of commercial product required to kill fifty per cent of a population of
rats.
12
energy is needed in inputs per a kcal of solar energy fixed by the photosynthesis
performed by crops).
Socio-economic variables
Finally, with the purpose of characterizing the eco-efficiency of our sample of farms,
the following structural variables have also been considered:
1. Farmer features: age, percentage of income derived from farming, level of education
(level of studies reached: school-leaving certificate, primary, secondary
and university), and specific professional training (type of agricultural instruction:
non-specialised agricultural training, extension courses, professional degrees and
university agricultural studies).
2. Farm features: farm surface area (in hectares) and percentage of farm surface
area subjected to the agri-environmental program.
4.3. Data gathering
We have relied on the data provided by an ad hoc survey as the main source of
primary information for the calculation of the variables mentioned above. With this
purpose, a specific questionnaire was designed including farmer features, as well as
farm characteristics and relevant information concerning productive processes
(technology and inputs used for every crop grown).
The universe of this survey was the 3,960 farms that, according to the last Agricultural
Census, operate in Campos County. Given the difficulties to implement random
sampling, stratified sampling was chosen based on the affiliation of these producers
to farmers’ unions (ASAJA, UPA and COAG). The survey was completed by personal
interviews during the period where farmers went to union offices in order to fill in forms
to obtain CAP subsidies and payments (March-April, 2008). Following this procedure,
171 valid questionnaires (farms) were finally obtained.
The primary information supplied by the survey has been complemented with secondary
additional information in order to calculate the variables required for the empirical
analysis (value added and the different environmental pressures), as has been
explained above. This information has been collected from different sources: a) scientific
literature for technical coefficients (Nitrogen inputc, Nitrogen outputc, Phosphorus
inputc, Phosphorus outputc, Energy inputsc, Energy outputsc, Quantity commercial
productmi and Lethal dosis50m) required to compute the environmental pressures considered,
b) official statistics for input and output prices required to calculate Salesk
and Intermediate costsk, and c) legal documents for CAP subsidies (Coupled subsidi-
13
esk) and agri-environmental payments (Agri-environmental paymentsk) received by
farmers. Table 1 depicts a summary descriptive analysis of these variables.
Table 1. Descriptive statistics. About here
5. RESULTS AND POLICY IMPLICATIONS
5.1. Assessing radial and pressure-specific eco-efficiency
This section presents and discusses our eco-efficiency estimates. On the one hand,
scores of radial eco-efficiency representing the proportional potential reduction of all
environmental pressures while maintaining value added have been computed by
solving expression (5) for each farm in the sample. On the other hand, pressurespecific
scores of eco-efficiency by farm and pressure level have been calculated
according to expression (9), using the information about pressure excesses previously
computed from expression (6). Table 2 presents some descriptive statistics for both
radial and pressure-specific measures of eco-efficiency, in addition to a measure of
the importance of slacks computed using expression (10).
Table 2. Computed scores of eco-efficiency. About here
Concerning radial scores of eco-efficiency, our results suggest that, on average,
the farmers in the sample could reduce their environmental pressures equiproportionally
by 44%, while maintaining their levels of value added, i.e., the average for radial
scores of eco-efficiency is 0.560. Pressure-specific eco-efficiency average scores are,
as they should be by construction, smaller than the average of radial eco-efficiency.
Sample averages for specialisation, nitrogen balance, phosphorus balance, risk of
pesticides and energy balance are 0.551, 0.437, 0.533, 0.535 and 0.544, respectively.
To illustrate the interpretation of these indicators of eco-efficiency, let us take farm
number 101 in our sample and environmental pressure nitrogen balance as an example.
This farm has a balance of nitrogen of 24.5 kg per hectare. According to its computed
score of radial eco-efficiency, which is equal to 0.700, it could reduce its pressure
on the environment by 30% while maintaining value added, which implies a potential
reduction of 7.3 kg per hectare. Additionally, the computed excess of nitrogen
specific pressure for this farm would allow for a further reduction of 4.4 kg per hectare.
Therefore, adding up radial reduction and pressure-specific excess, the aggregate
reduction in the balance of nitrogen necessary to achieve eco-efficiency amounts to
11.7 kg per hectare, such that the efficient pressure would be 12.8 kg per hectare.
Accordingly, the pressure-specific score of eco-efficiency for this farm is 0.521, which
stems from the comparison of eco-efficient pressure to actually observed environmental
pressure.
14
These outcomes show, in the first place, that farmers in the sample are rather ecoinefficient.
Secondly, eco-inefficiencies affect all the environmental pressures considered
in the analysis in a similar fashion, as differences across pressures are relatively
small. The greatest eco-inefficiency is observed in the balance of nitrogen, i.e., the
average maximum attainable reduction of this pressure while maintaining value
added stands at 56.3%, while the most eco-efficient management corresponds to
specialisation, with an average potential reduction close to 45%.
As regards the importance of slacks in explaining the aggregate potential reduction
of environmental pressures, our results show that their weight goes from 33% in the
case of the nitrogen balance to barely 3% for pressure specialisation. In addition, all
eco-inefficient farms in the sample have at least one slack in some pressure direction.
Although these figures do not suggest slacks are overly relevant, computing pressurespecific
measures of eco-efficiency, instead of a single radial measure, improves the
assessment of eco-efficiency of farming in the rain-fed agricultural system of Campos
County.
From a society perspective, it is worth analysing why farmers are so eco-inefficient,
as revealed by the foregoing results. In this sense, three main reasons can be given:
1. Technical inefficiency. If a farmer does not manage inputs efficiently from a
technical perspective, i.e., the amount of output can be maintained even if the
use of inputs is reduced, he cannot be eco-efficient. This fact has already been
demonstrated by Picazo-Tadeo and Reig-Martínez (2006a and 2007), who show
how farmers overuse inputs (nitrogen in their case studies), mostly because of inefficient
management, and how environmental pressures (non-point pollution in
their case studies) could be reduced by merely promoting best farming techniques.
In order to ascertain whether the same occurs in this case, a standard
technical efficiency analysis has also been performed2. This analysis reports an
average input-oriented technical efficiency of 0.814 for the sample of farms considered3.
Furthermore, technical efficiency has turned out to be highly correlated
with radial eco-efficiency (Pearson’s correlation coefficient of 0.66) and pressurespecific
eco-efficiency (Pearson’s correlation coefficients are 0.62, 0.43, 0.51,
0.56 and 0.64 for specialisation, nitrogen balance, phosphorus balance, risk of
pesticides and energy balance, respectively)4. Thus, we may confirm that also in
2 More specifically, an input-oriented DEA analysis with the objective of maximizing the ratio between
farms’ income –the sum of sales, coupled subsidies and agri-environmental payments– and a weighted
sum of the physical amounts of inputs they use, both variable – seeds, nitrogen and phosphorous fertilizers,
pesticides, labour, capital and energy– and fixed –land.
3 Other descriptive statistics are standard deviation 0.124, maximum 1 and minimum 0.467.
4 Furthermore, eleven out of sixteen farms scoring one for radial eco-efficiency also scored one for technical
efficiency.
15
this case technical inefficiency (inadequate farm management) is a major
source of eco-inefficiency.
2. Eco-efficiency as an environmental externality. As pointed out above, farmers
are private agents that are assumed to maximize their personal utility. Thus, they
make decisions ignoring every consequence not taken into account within their
personal utility function. This is the case of most environmental pressures. These
pressures must be considered as externalities; any change in these pressures affects
other agents’ utility without any consequence for the farmer (no compensation
is required or paid)5. This explains why no civic-minded behaviour should
be expected from these producers in order to minimize these environmental
pressures (improve their eco-efficiency level and provide further positive externalities).
As it is well known (Hodge, 2000), there are two main options to cope
with this market failure. The first one is to internalise the provision of further positive
externalities through economic instruments such as agri-environmental programs.
This involves compensating farmers for implementing environmentally friendly
techniques and practices (reduction of environmental pressures). The second
one is to impose stricter environmental standards through the implementation of
environmental regulations (command and control option) in order to make decreasing
said pressures compulsory.
3. Multi-criteria framework for farmers’ decision-making. A wide number of studies
reject the hypothesis that farmers seek to maximize profits or any utility function
with a single attribute, arguing that producers seek to optimise a broader set of
objectives such as the maximization of leisure time, the minimization of management
complexity, the minimization of working capital, etc. In this line, the research
by Gasson (1973), Harper and Eastman (1980), Cary and Holmes (1982),
Willock et al. (1999), or Gómez-Limón et al. (2004) is worth mentioning. The implication
is clear: the use of inputs by farmers must be explained, not only on the
basis of profit maximization, as considered in our (eco-)efficiency analysis, but
also from their respective utility functions in a multi-criteria context. This means
that part of farmers’ inefficient performance can be rationally explained consid-
5 Great difficulty arises from the general confusion regarding the matter of whether particular environmentally-friendly
agricultural techniques and practices provide benefits (generate positive externalities)
or just prevent harm (avoid negative externalities). It depends on who owns the environmental
property rights. If these rights belong to farmers, these techniques and practices are benefit providers. If
the public (the whole society) is the owner, they just prevent harm (Bromley and Hodge, 1990; Whitby,
2000; Ortiz and Estruch, 2004). In this sense, it is worth commenting that for all rain-fed farms considered
in the sample, the former case (farmers are the owner of these environmental property rights) applies.
This is because all farms fulfil CAP conditionality requirements, the environmental pressures generated
being below the legal standards. Thus, increasing eco-efficiency (decreasing pressures) must be considered
a positive externality generated by these farms.
16
ering that they wish to optimize a wider range of objectives within a multi-criteria
framework.
Although it is obvious that the three above-mentioned issues can explain farmers’
eco-inefficient behaviour, unfortunately, at least as far as we are aware, no methodological
approach is currently available to calculate the relative importance of each
one of them.
5.2. Can we explain farming eco-efficiency?
In the empirical literature in the field of efficiency measurement, it has been common
to perform two-stage analyses to investigate, in the second stage, the determinants
of efficiency scores obtained in the first stage. In performing the second stage,
a common practice has been to use regression analyses, mostly censored Tobit regression,
to test for the relationship between efficiency and some covariates. Nonetheless,
Simar and Wilson (2007) showed that second-stage analyses based on regressing
first-stage DEA efficiency estimates against a set of explanatory variables
might lead to inaccurate results, mainly because of the serial correlation of the firststage
DEA estimates and the correlation between the error term and the set of covariates
in the second stage. Instead, truncated regression and bootstrapping procedures
are proposed to allow for better estimation and statistical inference.
In our manuscript, a number of hypotheses concerning the determinants of ecoefficiency
have been tested using truncated regression analysis and confidence intervals
computed according to the single bootstrapping procedure proposed by Simar
and Wilson (2007: 41-42). In order to accommodate our analysis to the lefttruncated
distribution functions developed in this paper, in our second-stage regressions,
the dependent variable, i.e., the variable to be explained, is the inverse of the
first-stage DEA estimates of eco-efficiency6. This transformed variable ranges, in consequence,
from one to infinity7. That said, explaining farmers’ eco-inefficiency requires
following the next three steps:
Step 1. Use maximum likelihood to obtain estimates of , and  in the truncated
regression of the eco-inefficiency scores estimated in the DEA-based first
6 A similar transformation has been used in Picazo-Tadeo and García-Reche (2007) and Picazo-Tadeo et
al. (2009).
7 The following example illustrates the implications of this transformation in interpreting our results. Let us
assume that the score of eco-efficiency computed for, say, farm k’ is equal to 0.5. This means that this
farm could maintain its added value generating only fifty percent of its current level of environmental
pressures. The inverse of this score, i.e., the variable to be used as dependent variable in our second
regression analysis, is equal to 2, indicating that the farm is generating two times the environmental
pressures that it should generate if it were eco-efficient. In consequence, the larger the value of our
transformed variable the smaller the eco-efficiency (or the greater the eco-inefficiency).
17
stage (Eco-ineff) on a set of covariates zi, using the subset of i = 1,…,I < 171
eco-inefficient observations.8 Formally:
i i iEco-ineff z   (17)
Step 2. Loop over steps (2.1) to (2.3) L times to obtain a set of bootstrap estimates
for the parameters  and :
Step 2.1 For each i = 1,…,I, draw i from the following normal distribution:
   2 ˆ0, left truncated at point 1ˆ  iN z , (18)
where ˆ are the estimates of  obtained from regression (17).
Step 2.2 Yet again, for each i = 1,…,I, compute:
*
i i i
ˆEco-ineff z   (19)
Step 2.3 Use the maximum likelihood method to estimate the following
truncated regression:
*
i i iEco-ineff z   (20)
Jointly, steps (2.1) to (2.3) yield a set of bootstrap estimates of  and :
  
L
b b
B

 * *
1
ˆ , ˆ  (21)
Step 3. Finally, use values in B and the original estimates of  and  to construct estimated
confidence intervals for these parameters.
Given that in the assessment of performance carried out in the fist-stage of our research
we found that differences of eco-efficiency across environmental pressures
are not important, in the second-stage we explain the radial scores of eco-efficiency.
Furthermore, the number of replications in the bootstrap procedure has been set
equal to 1,000. Table 3 displays the estimated parameters and their confidence intervals
using habitual confidence levels. Regarding the features capable of influencing
eco-efficiency, we include traditional variables such as farmer educational level,
specific professional training, age and farm size, in addition to variables representing
the percentage of income derived from farming and the percentage of farm surface
area subjected to agri-environmental programs. Details on how these variables have
8 According to the first-stage results, 155 farms in the sample are found to be eco-inefficient.
18
been constructed are in Section 4.2.
Table 3. Truncated regression. Bootstrapped confidence intervals. About here
Empirical evidence reveals that the level of educational reached by farmers affects
their eco-efficiency. With a level of confidence of 95%, it can be stated that
farmers with secondary studies are less eco-efficient than farmers with university studies,
which is the category omitted in the truncated regression. However, the variable
primary studies is not statistically significant at standard confidence levels. Furthermore,
the percentage of farm surface area that benefits from agri-environmental
programs exercises a significant effect on eco-efficiency, the larger the percentage
the higher the eco-efficiency. Aside from these relationships, according to our results
none of the remaining variables seem to affect the eco-efficiency of the farmers in
the sample.
From these results, three main conclusions can be drawn:
1. Eco-efficiency can hardly be explained through traditional socio-structural features
(farmers’ age, income, etc.). Thus, it is probable that it depends on farmers’
psychological traits, which are more difficult to observe and quantify (environmental
concerns, agricultural vocation, etc.), or other productive characteristics
not considered in this analysis (level of outsourcing, etc.). Thus, further research in
this direction would be useful. In addition, it would be also interesting to explore
the explanatory factors behind the three main sources of eco-inefficiency
pointed out above (technical efficiency, farmers’ consideration of environmental
externalities and multi-criteria decision-making).
2. The level of education was, as already noted, the only significant sociodemographic
variable. The results obtained confirm that farmer education improves
eco-efficiency, as has already been proved in the literature (see for example,
Phillips, 1994 or Thiam et al. 2001). In any case, it is worth noting that these
results hide two different realities regarding farmers’ profile. First, some of these
farmers with university studies have a degree in agriculture. In these cases, it can
be easily assumed that their higher level of technical knowledge allows them to
run their farms more eco-efficiently. However, at the same time, there is a large
percentage of farmers with university studies that are not directly related to farming
activities; they are lawyers, physicians, teachers… In these cases, no especial
knowledge or training about agriculture can be expected. For most of them,
farming is just a secondary source of income, carried out in non-professional
fashion by contracting-out most farming chores. The eco-efficient performance
of these farmers can be explained by the relative advantage of managerial
19
strategy based on outsourcing. In fact, as Picazo-Tadeo and Reig-Martínez
(2006b) have shown in other Spanish agricultural systems, outsourcing labour and
capital (through contracts with agricultural service firms and co-operatives) also
allows farmers to improve their efficiency.
3. Finally, it should be stressed that agri-environmental programs seem to improve
farms’ eco-efficient performance, being an effective policy to reduce environmental
pressures stemming from farming sector. Theoretically, agri-environmental
payments obtained by those farmers subscribing these contracts are equivalent
to the additional costs needed to fulfil program requirements that go beyond
the CAP single payment’s conditionality requested from all producers. Thus, it
must be assumed that the economic value added per hectare remains constant
for these farms, and the increase in eco-efficiency responds to changes in practices
and techniques implemented in order to fulfil environmental requirements.
However, some doubts arise regarding the efficiency of these programs (costbenefit
balance compared with other policy options). Taking into account the
high level of eco-inefficiency of these farmers, is it really necessary to spend public
resources to reduce environmental pressures? As our results suggest, these
pressures could be cut by merely adopting better management techniques and
practices that would not affect farmers’ profits (and where no compensation is
needed). Thus, command and control measures, i.e., increasing conditionality
requirements to obtain CAP payments, could be a better option. Nevertheless,
further research into this issue is also required.
6. SUMMARY AND CONCLUSIONS
Sustainable development is a matter of concern that has received increasing attention
since the 1980s from both policy-makers and academics. Furthermore, in the last
few years, researchers in the fields of economics and ecology have shown mounting
interest in assessing ecological-economic efficiency, more commonly known as ecoefficiency.
Measuring eco-efficiency is important as it might provide policy-makers
with valuable information for designing polices aimed at achieving sustainable development.
Starting from the most common definition of eco-efficiency as a ratio between
economic value added and environmental pressures, this paper assesses ecoefficiency
at micro-farm level using a sample of Spanish farmers operating in the rainfed
agricultural system of Campos County. Five environmental pressures, namely,
specialisation, nitrogen and phosphorus balances, pesticide risk and energy balance,
are considered in the analysis. A score of eco-efficiency for each farm in the sample,
as well as pressure-specific scores of performance, also at farm level, are computed
using Data Envelopment Analysis techniques. Furthermore, the determinants of eco-
20
efficiency are investigated using truncated regression and bootstrapping techniques.
Our major findings are the following. On the one hand, concerning the assessment
of eco-efficiency, the farmers in the sample are observed to be highly eco-inefficient.
In addition, differences in eco-efficiency among environmental pressures are found to
be relatively small, the most eco-inefficient behaviour corresponding to the pressure
stemming from the use of the input nitrogen. At least three reasons stand behind these
results: the existence of inefficient practices among farmers concerning the technical
management of inputs, the lack of consideration on behalf of farmers for environmental
externalities and, finally, the assumption of a multicriteria framework for farmer
decision-making. Although, as far as we know, no method currently exists to estimate
the quantitative contribution of these factors to eco-inefficiency, indirect empirical
evidence is found pointing out the explanatory relevance of technical inefficiencies.
On the other hand, and regarding the determinants of eco-efficiency, we find
that it can barely be explained by conventional socio-economic variables. Only the
education variable is significant when it comes to explaining eco-efficiency. Farmers
with secondary studies are less eco-efficient than farmers that have completed university
studies. Furthermore, we find that CAP agri-environmental programs seem to
be an effective policy to improve farms’ eco-efficiency.
All these results provide some relevant policy implications to better integrate farming
activities and the environment. In this sense, it has been shown that there are several
ways to improve agricultural eco-efficiency. First, taking into account the close
relationship between technical efficiency and eco-efficiency, further public expenditure
in agricultural extension and farmer training could be rewarding. Second, agrienvironmental
policy should also be strengthened to avoid market failures in the provision
of externalities. For this purpose, it has been demonstrated that a wider implementation
of agri-environmental programs can be effective in decreasing environmental
pressures derived from the farming sector. However, it has been also verified
that a more efficient way to achieve the same objective is by setting up stricter conditionality
requirements to obtain CAP subsidies. This measure does not require any
public expenditure, except monitoring costs related to the public control needed to
enforce these requirements, and does not necessarily affect farmers’ profitability. Furthermore,
it is worth commenting that all this policy advice should be taken with care
as the results of implementing them could be affected by differences in farmers’ multicriteria
utility functions. Such differences could lead to unexpected productive decisions
(crop plan, inputs use,…) and, thus, to unforeseen policy results.
Finally, let us remark that, although the design of effective environmental policies
involves manifold considerations, assessing eco-efficiency might help policy-makers to
21
design agricultural policies more capable of achieving the general objective of agricultural
sustainability and, particularly, the sustainability of specific agricultural systems.
Nonetheless, further research is still needed in several areas. First, in order to
quantify the relative importance of the different issues behind eco-inefficiency. Only
in this way can a better founded policy response be provided. Second, other farmer
features including psychological aspects such as environmental concerns should be
considered in the future to explain eco-efficiency. Finally, more research is also required
in order to obtain a more precise assessment of efficiency in terms of the costs
and benefits of CAP agri-environmental programs.
REFERENCES
Ali, A. and Seiford, L. ‘The mathematical programming approach to efficiency analysis’.
In Fried, H., Lovell, C.A.K. and Schmidt, S. (eds.), The measurement of productive
efficiency: Techniques and applications (Oxford: Oxford University Press, 1993,
pp.120-159).
Allen, K. ‘DEA in the ecological context–an overview’, in G. Westermann (ed.), Measuring
the Efficiency in the Private and Public Service Sector (Wiesbaden: Gabler,
1999, pp. 203-235).
Bell, S. and Morse, S. Sustainability indicators. Measuring the incommensurable? (London:
Earthscan, 2008).
Böhringer,C. and Jochem, P.E.P. ‘Measuring the immeasurable – A survey of sustainability
indices’, Ecological Economics, Vol. 63, (2007) pp. 1-8.
Bromley, D.W. and Hodge, I. ‘Private property rights and presumptive policy entitlements:
reconsidering the premises of rural policy’, European Review of Agricultural
Economics, Vol. 17, (1990) pp. 197-214.
Cary, J.W. and Holmes, W.E. ‘Relationships among farmers’ goals and farm adjustment
strategies: some empirics of a multidimensional approach’, The Australian
Journal of Agricultural Economics, Vol. 26, (1982) pp. 114-130.
Charnes, A., Cooper, W.W. and Rhodes, E. ‘Measuring the efficiency of decision making
units’, European Journal of Operational Research, Vol. 2, (1978) pp. 429-444.
Cook, W.D. and Seiford, L.M. ‘Data envelopment analysis (DEA) – Thirty years on’,
European journal of Operational Research, Vol. 192, (2009) pp. 1-17.
Cooper, W.W., Seiford, L.M. and Tone, K. Data Envelopment Analysis. A comprehensive
text with models, applications, references and DEA-Solver Software (New York:
Springer, 2007).
22
De Koeijer, T.J., Wossink, G.A.A., Struik, P.C. and Renkema, J.A. ‘Measuring agricultural
sustainability in terms of efficiency: the case of Dutch sugar beet growers’, Journal
of Environmental Management, Vol. 66, (2002) pp. 9-17.
EEA, European Environment Agency. Agriculture and environment in EU-15 – the IRENA
indicator report. Report No. 6/2005 (Copenhagen: EEA, 2005).
Ekins, P. ‘Eco-efficiency. Motives, drivers and economic implications’, Journal of Industrial
Ecology, Vol. 9, (2005) pp. 12-14.
Farrell, M.J. ‘The measurement of productive efficiency’, Journal of the Royal Statistical
Society, Series A, General, Vol. 120, (1957) pp. 253-281.
Gasson, R. ‘Goals and values of farmers’, Journal of Agricultural Economics, Vol. 24,
(1973) pp. 521-537.
Gómez-Limón, J.A. and Sanchez-Fernandez, G. ‘Empirical evaluation of agricultural
sustainability using composite indicators’, Ecological Economics, (2010) in press.
Gómez-Limón, J.A., Riesgo, L. and Arriaza, M. ‘Multi-Criteria analysis of input use in agriculture’,
Journal of Agricultural Economics, Vol. 55, (2004) pp. 541-564.
Harper, W.H. and Eastman, C. ‘An evaluation of goal hierarchies for small farm operations’,
American Journal of Agricultural Economics, Vol. 62, (1980) pp. 742-747.
Hodge, I. ‘Agri-environmental relationships and the choice of policy mechanism’, The
World Economy, Vol. 23, (2000) pp. 257-273.
Huppes, G. and Ishikawa, M. ‘A Framework for Quantified Eco-efficiency Analysis’,
Journal of Industrial Ecology, Vol. 9, (2005) pp. 25-41.
Kallas, Z., Gómez-Limón, J.A. and Barreiro-Hurlé, J. ‘Decomposing the value of agricultural
multifunctionality: Combining contingent valuation and the analytical hierarchy
process’, Journal of Agricultural Economics, Vol. 58, (2007) pp. 218-241.
Koopmans, T. ‘Analysis of production as an efficient combination of activities’, in T.
Koopmans (ed.) Activity analysis of production and allocation (New York: John
Wiley & Sons, 1951, pp. 33-97).
Korhonen, P.J. and Luptacik, M. ‘Eco-efficiency analysis of power plants: An extension
of data envelopment analysis’, European Journal of Operational Research, Vol.
154, (2004) pp. 437-446.
Kortelainen, M. and Kuosmanen, T. Measuring eco-efficiency of production: a frontier
approach, EconWPA, Working Paper 0411004 (St. Louis: Washington University,
2005) .
Kuosmanen, T. and Kortelainen, M. ‘Measuring Eco-efficiency of Production with Data
Envelopment Analysis’, Journal of Industrial Ecology, Vol. 9, (2005) pp. 59-72.
23
OECD, Organization for Economic Co-operation and Development. Eco-efficiency
(Paris: OECD, 1998).
OECD, Organization for Economic Co-operation and Development. Environmental
indicators for agriculture. Volume 3 – Methods and Results (Paris: OECD, 2001).
Ortiz, D. and Estruch, V. ‘The role of agri-environmental measures in the definition of
property rights’, in G. Van Huylenbroeck, W. Verbeke and L. Lauwers (eds.) Role of
institutions in rural policies and agricultural markets (Amsterdam: Elsevier, 2004, pp.
335-348).
Phillips, J.M. ‘Farmer Education and Farmer Efficiency: A Meta-Analysis’. Economic
Development and Cultural Change, Vol. 43, (1994) pp. 149-165.
Picazo-Tadeo, A.J. and García-Reche, A. ‘What makes environmental performance
differ between firms? Empirical evidence from the Spanish tile industry’, Environment
and Planning A, Vol. 39, (2007) pp. 2232-2247.
Picazo-Tadeo, A.J. and Reig-Martínez, E. ‘Agricultural externalities and environmental
regulation: evaluating good practice in citrus production’, Applied Economics,
Vol. 38, (2006a) pp. 1327–1334.
Picazo-Tadeo, A.J. and Reig Martínez, E. ‘Outsourcing and efficiency: the case of
Spanish citrus farming’, Agricultural Economics, Vol. 35, (2006b) pp. 213-222.
Picazo-Tadeo, A.J. and Reig-Martínez, E. ‘Farmers' costs of environmental regulation:
Reducing the consumption of nitrogen in citrus farming’, Economic Modelling, Vol.
24, (2007) pp. 312-328.
Picazo-Tadeo, A.J., Reig-Martínez, E. and Estruch, V. ‘Farming efficiency and the survival
of valuable agro-ecosystems: A case study of rice farming in European Mediterranean
wetlands’, Open Environmental Sciences, Vol. 3, (2009) pp. 69-78.
Rigby, D., Howlett, D. and Woodhouse, P.A. Review of Indicators of Agricultural and
Rural Livelihood Sustainability. Working Paper 1. Research Project No.R7076CA
(Manchester: University of Manchester, Department for International Development,
2000).
Schaltegger, S. and Sturm, A. ‘Ökologische Rationalität’, Die Unternehmung, Vol. 4,
(1990) pp. 273-290.
Schaltegger, S. Corporate Environmental Accounting (Chichester: John Wiley and
Sons Ltd., 1996).
Scheel, H. ‘Undesirable outputs in efficiency valuations’, European Journal of Operational
Research, Vol. 132, (2001) pp. 400-410.
24
Schidemeny, S. with the Business Council for Sustainable Development (BCSD).
Changing course: A global business perspective on development and the environment
(Cambridge, MA: MIT Press, 1992).
Simar, L. and Wilson P.W. ‘Estimation and inference in two-stage, semi-parametric
models of production processes’, Journal of Econometrics, Vol. 136, (2007) pp. 31-
64.
Thiam, A., Bravo-Ureta, B.E., Rivas. T.E. ‘Technical efficiency in developing country agriculture:
a meta-analysis’. Agricultural Economics, Vol. 25, (2001) pp. 235-243.
Torgersen, A., Førsund, F. and Kittelsen, S. ‘Slack-adjusted efficiency measures and
ranking of efficient units’, Journal of Productivity Analysis, Vol. 7, (1996) pp. 379-98.
Tyteca, D. ‘On the Measurement of the Environmental Performance of Firms – A Literature
Review and a Productive Efficiency Perspective’, Journal of Environmental
Management, Vol. 46, (1996) pp. 281-308.
Van Cauwenbergh, N., Biala, K., Bielders, C., Brouuckaert, V., Franchois, L., Cidad,
V.G., Hermy, M., Mathijs, E., Muys, B., Reinjders, J., Sauvenier, X., Valckx, J., Vanclooster,
M., Van der Veken, B., Wauters, E. and Peeters, A. ‘SAFE – a hierarchical
framework for assessing the sustainability of agricultural systems’, Agriculture, Ecosystems
and Environment, Vol. 120, (2007) pp. 22-242.
Van der Werf, H.M.G. and Petit, J. ‘Evaluation of the environmental impact of agriculture
at the farm level: a comparison and analysis of 12 indicator-based methods’,
Agriculture, Ecosystems and Environment, Vol. 93, (2002) pp. 131-145.
WBCSD, World Business Council for Sustainable Development. Measuring ecoefficiency.
A guide to reporting company performance (Geneva: WBCSD, 2000).
WCED, World Commission on Environment and Development. Our Common Future.
The Brundtland Report (Oxford: Oxford University Press, 1987).
Whitby, M. ‘Challenges and options for the UK agri-environment’. Journal of Agricultural
Economics, Vol. 51, (2000) pp. 17-32.
Willock, J., Deary, I.J., Edwards-Jones, G., Gibson, G.J., McGregor, M.J., Sutherland, A.,
Dent, J.B., Morgan, O. and Grieve, R. ‘The role of attitudes and objectives in farmer
decision-making: business and environmentally-oriented behaviour in Scotland’,
Journal of Agricultural Economics, Vol. 50, (1999) pp. 286-303.
Zhang, B., Bi, J., Fan, Z., Yuan, Z. and Ge, J. ‘Eco-efficiency analysis of industrial system
in China: A data envelopment analysis approach’, Ecological Economics, Vol. 68,
(2008) pp. 306-316.
25
Zhou, P., Ang, B.W. and Poh, K.L. ‘A survey of data envelopment analysis in energy
and environmental studies’, European Journal of Operational Research, Vol. 189,
(2008) pp. 1-18.
26
Table 1a
Sample description (continuous variables)
Mean
Standard
deviation Maximum Minimum
OUTPUT
Sales (€ per hectare) 471.0 80.9 892.4 299.0
Coupled subsidies (€ per hectare) 34.7 5.6 39.0 0.0
Agri-environmental payments (€ per hectare) 16.4 29.7 195.9 0.0
INPUTS
Seeds (€ per hectare) 55.4 11.7 106.0 34.1
Nitrogen (€ per hectare) 64.1 39.6 195.7 0.0
Phosphorus (€ per hectare) 23.3 11.4 84.0 0.0
Pesticides (€ per hectare) 21.1 22.8 221.7 0.0
Energy (€ per hectare) 38.9 10.1 81.1 18.4
ENVIRONMENTAL PRESSURES
Specialisation (%) 64.9% 0.210 100.0% 25.0%
Nitrogen balance (kg N per hectare) 21.4 28.2 196.9 0.0
Phosphorus balance (kg P per hectare) 31.0 46.5 494.9 0.0
Pesticides risk (kg rat per hectare) 0.639 0.573 4.764 0.000
Energy balance (%) 301.9% 0.064 476.0% 136.3%
SOCIO-ECONOMIC VARIABLES
Age (years) 44.3 9.4 65.0 23.0
Income coming from agriculture (%) 85.0% 0.26 100.0% 0.0%
Land (hectare) 122.0 117.1 820.0 1.0
Surface subjected to agri-environmental
payments (%) 19.8% 0.318 94.8% 0.0%
Number of farms 171
27
Table 1b
Sample description (categorical variables)
Frequency
SOCIO-ECONOMIC VARIABLES
Education
School-leaving certificate 18.1%
Primary school 39.8%
Secondary school 33.9%
University 8.2%
Agricultural training
None 12.9%
Basic: agricultural extension 74.3%
Medium: professional training 8.8%
Upper: university 4.0%
Number of farms 171
28
Table 2
Computed scores of eco-efficiency
Mean
Standard
deviation Maximum Minimum
Importance
of slacks (%)
Radial eco-efficiency 0.560 0.215 1.000 0.178 Pressure-specific
eco-efficiency
Specialisation 0.551 0.217 1.000 0.178 2.8%
Nitrogen balance 0.437 0.334 1.000 0.001 33.0%
Phosphorus balance 0.533 0.276 1.000 0.009 21.0%
Pesticides risk 0.535 0.242 1.000 0.079 15.1%
Energy ratio 0.544 0.216 1.000 0.178 3.5%
29
Table 3
Truncated regression. Bootstrapped confidence intervals. The dependent variable is the inverse of radial eco-efficiency scores.
99% confidence 95% confidence 90% confidence
Estimated
parameter
Lower
bound
Upper
bound
Lower
bound
Upper
bound
Lower
bound
Upper
bound
Age (years) 0.0002 -0.0250 0.0282 -0.0209 0.0224 -0.0175 0.0185
Land (hectare) 0.1e-4 -0.0018 0.0026 -0.0017 0.0021 -0.0014 0.0018
Income coming from agriculture (%) 0.0187 -0.9798 0.8760 -0.8273 0.7554 -0.6661 0.6365
Surface subjected to agri-environmental payments (%) -1.5730 -2.5341 -0.4737 -2.3796 -0.6807 -2.2917 -0.9136
Education: school-leaving certificate and primary school a 0.6010 -0.7484 1.6322 -0.5173 1.4629 -0.3240 1.3268
Education: secondary school a 1.1068 -0.2842 2.1085 0.0024 1.9693 0.1712 1.8591
Agricultural training: basic, medium and upper b 0.3028 -0.5921 1.0250 -0.4016 0.9292 -0.2709 0.8565
Constant 1.0191 -0.6266 2.9701 -0.4374 2.5785 -0.2200 2.3103
Sigma 0.9897 0.7592 1.2283 0.8304 1.1972 0.8615 1.1798
Number of observations 171
a Concerning education, the category omitted is University.
b In this case, the category omitted is no specialised agricultural training.