Environmental and Resource Economics
Eco-efficiency among dairy farmers: The importance of socio-economic characteristics
and farmer attitudes
--Manuscript Draft--
Manuscript Number: EARE-D-13-00422
Full Title: Eco-efficiency among dairy farmers: The importance of socio-economic characteristics
and farmer attitudes
Article Type: Manuscript
Keywords: Eco-efficiency; Dairy farms; Socioeconomic Variables; Data Envelopment Analysis
Abstract: The aim of this paper is to assess the eco-efficiency of dairy farms in Spain. To do so,
we use data from a survey carried out in 2010 for the specific purpose of analysing the
environmental performance of 50 dairy farms in the Spanish region of Asturias. The
survey contains information on nutrients balances and greenhouse gas emissions
which is used to calculate environmental pressure indicators. Eco-efficiency is
measured using Data Envelopment Analysis (DEA). We analyse the influence of the
socio-economic characteristics and farmers' attitudes in explaining these eco-efficiency
scores using truncated regression and bootstrapping procedures. On average, the
dairy farms are found to to be highly eco-inefficient. Among our results, farmers that
are younger, that plan to continue in operation in the foreseeable future, and that have
greater participation in training schemes are found to be more eco-efficient. Selfreported
positive environmental habits are also reflected in actual eco-efficient
performance.
Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Environmental & Resource Economics manuscript No.
(will be inserted by the editor)
Eco-eﬃciency among dairy farmers: The importance of
socio-economic characteristics and farmer attitudes
Abstract The aim of this paper is to assess the eco-eﬃciency of dairy farms in Spain.
To do so, we use data from a survey carried out in 2010 for the speciﬁc purpose of
analysing the environmental performance of 50 dairy farms in the Spanish region of Asturias.
The survey contains information on nutrients balances and greenhouse gas emissions
which is used to calculate environmental pressure indicators. Eco-eﬃciency is measured
using Data Envelopment Analysis (DEA). We analyse the inﬂuence of the socioeconomic
characteristics and farmers’ attitudes in explaining these eco-eﬃciency scores
using truncated regression and bootstrapping procedures. On average, the dairy farms
are found to to be highly eco-ineﬃcient. Among our results, farmers that are younger,
that plan to continue in operation in the foreseeable future, and that have greater participation
in training schemes are found to be more eco-eﬃcient. Self-reported positive
environmental habits are also reﬂected in actual eco-eﬃcient performance.
*Manuscript, excluding author names and affiliation
2
1. Introduction
The environmental sustainability of agricultural activity is an issue of increasing concern
to policymakers at national and international levels. At European level, Common
Agricultural Policy (CAP) reforms have been giving increasing priority to the environment
and future objectives point to the pursuit of economic growth in the agricultural
sector while preventing environmental damage. Within the agricultural sector, dairy
farming has not escaped scrutiny and the negative impacts of this activity on the environment
have been well documented. In particular, land, water, air, biodiversity and
the landscape can all be aﬀected as a consequence of, among others, the generation and
management of animal waste, the use of fertilizers and pesticides in the production of
fodder, and the emission of greenhouse gases (CEAS, 2000; COWI, 2001; OECD, 2004).
As examples, nitrate leaching and phosphorus run-oﬀ lead to eutrophication of the surface
water, evaporation of ammonia and leaching to groundwater, while methane and
nitrous oxide emissions from the dairy sector are signiﬁcant contributors to greenhouse
gas emissions (Basset-Mens et al., 2009). Concern over such eﬀects at international
level has led to policy responses such as the European Union’s 1991 EU Nitrates Directive
and its proposed Water Framework Directive, both of which aﬀect the dairy
sector, and the FAO has called for steps to minimize the production of environmental
pollutants from dairy farming, with special emphasis on both nutrients and greenhouse
gases (FAO and IDF, 2011).
The literature on assessing the environmental impacts of agricultural activities
frequently invokes the concept of eco-eﬃciency. Eco-eﬃciency has been deﬁned by the
OECD as “the eﬃciency with which ecological resources are used to meet human needs”
(OECD, 1998), and thus takes into account both the environmental and economic
3
objectives of farms. An ineﬃcient use of resources, particularly nutrients, can have
negative eﬀects on both the environment and the economic results of the production
system (Oenema and Pietrzak, 2002). The term eco-eﬃciency itself was coined by the
World Business Council for Sustainable Development (WBCSD) in their 1993 report
(Schmidheiney, 1993) and it is based on the concept of creating more goods and services
while using fewer resources, thereby generating less pollution.
While improvements in eco-eﬃciency do not guarantee sustainability per se insofar
as higher eco-eﬃciency scores are compatible with (potentially unsustainable) increases
in environmental pressures, Kuosmanen and Kortelainen (2005) point out that measurement
of eco-eﬃciency is critically important for at least two reasons. Firstly, an
improvement of eco-eﬃciency is often the most cost-eﬀective way of reducing environmental
pressures. Secondly, policies targeted at eﬃciency improvements are more easily
implemented than policies that restrict the level of economic activity. In sectors such
as dairy farming where pollution is a problem but proﬁt margins are becoming tighter
and tighter over time, it is of prime interest to policymakers to have information on
the extent - if any - to which producers could reduce environmental pressures while
maintaining (or improving on) their existing economic value added.
Production frontier models are a natural tool in measuring the eco-eﬃciency of
producers,. In such models, an eﬃcient frontier is estimated and the relative eﬃciency
of producers is measured as the distance from this frontier (Coelli et al., 2005; Fried
and Schmidt, 2008). Production frontier models have been widely used in the literature
to measure environmental performance and an early survey of the literature by
Tyteca (1996) discussed the usefulness of productive eﬃciency models in this context.
Of particular relevance to the present study are what Lauwers (2009) refers to as frontier
eco-eﬃciency (FEE) models. As these FEE models relate ecological and economic
4
outcomes rather than the conventional inputs and outputs used in standard production
eﬃciency models, Lauwers (2009) refers to them as the “frontier operationalisation”
of the traditional eco-eﬃciency concept. Among studies following this approach are
Callens and Tyteca (1999), Tyteca (1999), Kuosmanen and Kortelainen (2005), Kortelainen
(2007) and Picazo-Tadeo et al. (2011).
The nonparametric Data Envelopment Analysis (DEA) method has been frequently
used for evaluating producers performance in the presence of adverse environmental
impacts (F¨are et al., 1989; Ball et al., 1994). F¨are et al. (1989) modify the eﬃciency
measures proposed by F¨are et al. (1985) to allow for an asymmetric treatment of
desirable and undesirable outputs. Pollutants are treated as weakly (costly) disposable
outputs whereas desirable outputs are strongly (freely) disposable (F¨are et al., 1989).
In the literature on technical and environmental performance measurement, pollutants
are usually treated either as undesirable outputs (F¨are et al., 1989; Ball et al., 1994;
Pittman, 1983; Reinhard et al., 1999; Oude Lansink and Silva, 2003; Oude Lansink and
Bezlepkin, 2003) or undesirable inputs (Tyteca, 1997), although there is no diﬀerence
between the two approaches conceptually.
While eco-eﬃciency has been measured before using DEA in the agricultural sector
(Picazo-Tadeo et al., 2011), the literature is lacking an analysis of eco eﬃciency in dairy
farming. Furthermore, farmers’ socio-economic characteristics and attitudes towards
the environment have been shown to be important factors aﬀecting the participation
in agri-environmental schemes (Wynn et al., 2001; Defrancesco et al., 2008). However,
the impact of farmer’s attitudes on eco-eﬃciency has not been investigated so far.1
In the light of the foregoing, the objective of this paper is to measure eco-eﬃciency
of dairy farms and to determine the impacts of farmers’ socio-economic characteristics
and attitudes toward the environment on their eco-eﬃciency.
5
The empirical application focuses on a sample of 50 dairy farmers from Asturias
in Spain. To get improved estimates of eco-ineﬃciency and better inference on its determinants,
we use a two-stage double bootstrap DEA procedure proposed by Simar
and Wilson (2007). In the ﬁrst stage of this procedure, distance functions are used
to measure eco-eﬃciency scores, and in the second stage these scores are regressed on
covariates related to attitudes towards the environment and socio-economic characteristics.
Latruﬀe et al. (2008) applied this technique to explain technical eﬃciency in
livestock and crop farms in the Czech Republic. The same procedure was applied by
Weninger (2008) to study the technnical eﬃciency of a Mexican ﬁshery and by Olson
and Vu (2009) to study economic eﬃciency of farms in Minnesota. However, as far as
we are aware, the two-stage double bootstrap DEA model has not been used previously
for analysing the determinants of economic and environmental performance.
The paper proceeds as follows. In Section 2 we discuss the concept of eco-eﬃciency
and the methodology used to estimate eco-eﬃciency scores and the factors determining
them. Section 3 describes the data we use. The results are presented and discussed in
Section 4, and Section 5 concludes.
2. Eco-eﬃciency: Concept and methodology used for measurement
2.1. Measuring eco-eﬃciency
To operationalize the concept of eco-eﬃciency, we follow the approach described in Kuosmanen
and Kortelainen (2005) which deﬁnes eco-eﬃciency in the traditional way as a
ratio between economic value added and environmental damage. Environmental damage,
p, is measured by aggregating the K environmental pressures (p1,. . . ,pK) which
are induced by the production activity. Kuosmanen and Kortelainen (2005) use the
6
tools of production economics and their point of departure is a pressure-generating or
pollution-generating technology set T = {(v, p) R (1+K)
| v can be generated by p}.
The technology set describes all the feasible combinations of economic value, v, and
environmental damage, p.
The pressure-generating technology set and the concept of eco-eﬃciency are illustrated
in Figure 1 below for the simple case of two environmental pressures. Production
activity generates economic value added but also induces two environmental pressures,
p1 and p2. A producer is deemed to be eco-eﬃcient if and only if it is impossible to
reduce one of the pressures without simultaneously increasing the other or decreasing
economic value. The set of eco-eﬃcient combinations is represented by the eﬃcient frontier,
T. Points below the frontier are unfeasible, while points above it are eco-ineﬃcient.
In the ﬁgure, the combination of pressures represented by A, which generates economic
value added v, is clearly eco-ineﬃcient as both environmental pressures could be reduced
without reducing economic value added.
A measure of eco-ineﬃciency can be found by measuring the radial distance from
a point A to the eﬃcient frontier, and a farm’s eﬃciency score is given by the ratio
OE/OA which takes values less than or equal to 1, with a value of 1 implying ecoeﬃciency.
For an ineﬃcient combination such as A in Figure 1, OE/OA < 1. [INSERT FIGURE 1 ABOUT
HERE]To empirically measure eco-ineﬃciency for a sample of producers, we need to aggregate
individual environmental pressures into an environmental damage indicator. We
follow the Data Envelopment Analysis (DEA) approach proposed by Kuosmanen and
Kortelainen (2005) for aggregating environmental pressures. Eco-eﬃciency is deﬁned
as the ratio:
Economic value added
Environmental pressures
=
vi
D(pi)
(1)
7
where vi is the Economic value added indicator of farm i and D is the function that
aggregates the environmental pressures into a single environmental pressure indicator
by taking a linear weighted average of the individual environmental pressures, i.e.,
D(pi) = w1p1 + w2p2 + ... + wkpk where wk is the weight accorded to environmental
pressure k.
In order to identify weights wk, we use DEA as a nonsubjective weighting method, i.e.,
it determines the weights that maximize the eco-eﬃciency score of farm i belonging
to the sample of i = 1, ..., N farms. The DEA eco-eﬃciency score of farm i’ can be
computed from the following programming problem:
maximize
wni
Eco − eﬃciencyi =
vi
K
k=1 wki pki
subject to
vi
K
k=1 wki pki
≤ 1 i = 1, ...N,
wki ≥ 0 k = 1, ...K
(2)
The above formulation involves a non-linear objective function and non-linear constrains,
which is computationally diﬃcult. This problem can be linearized by taking
the inverse of the eco-eﬃciency ratio and solving the reciprocal problem (Kuosmanen
8
and Kortelainen, 2005; Picazo-Tadeo et al., 2011):
minimize
θi ,zi
Eco − eﬃciency−1
i’ = θi’
subject to
vi ≤
N
i=1
zivi,
θi pki ≥
N
i=1
zipki k = 1, ..., K,
zi ≥ 0 i = 1, . . . , N
(3)
where zi is a set of intensity variables representing the weight of each farm i in the
eco-eﬃcient frontier.
The DEA eco-eﬃciency score which solves this problem for farm i, θ∗
i , can be
interpreted as a distance to the eco-eﬃciency frontier. It indicates the maximum potential
equiproportional reduction in all environmental pressures that could be achieved
while maintaining the present level of economic activity. This corresponds to the ratio
OE/OA for a farm operating at point A in Figure 1. Clearly, the DEA score equals one
for an eco-eﬃcient farm, in which case point A would be on the frontier, with values
lower than one indicating eco-ineﬃciency. Obviously, the further the distance of the
farm from the frontier, the lower the eco-eﬃciency score and the greater the scope for
improvement in a farm’s environmental performance.
2.2. Explaining eco-eﬃciency: truncated regression
A typical approach in studies computing eﬃciency measures using DEA is to conduct
a second-stage regression analysis to relate these eﬃciency scores to a set of explanatory
variables, with a censored Tobit regression being the method most often applied.
9
However, Simar and Wilson (2007) show that using a Tobit regression to perform a
two-stage analysis to analyse the determinants of eﬃciency scores might be inappropiate
because serial correlation of the ﬁrst stage DEA eﬃciency estimates is not taken
into account. Therefore, we apply the second of two truncated regression and bootstrapping
procedures (Algorithm #2) proposed by Simar and Wilson (2007), which
improves on inference and performs bias correction.
In order to follow the methodology described in Simar and Wilson (2007), we need a
distribution with left-truncation at 1, so we transform the variable to be explained in
the second stage by taking the inverse of the eco-eﬃciency scores estimated in the ﬁrst
stage according to (3).2
The truncated regression of the eco-eﬃciency scores on a set
of explanatory variables can be written as:
Σi = ziβ + i ≥ 1 (4)
where Σi represents (the inverse of) the eﬃciency scores.
Bootstrapping leads to consistent estimates of β. The following is a description of
the bootstrap procedure using Algorithm #2 of Simar and Wilson (2007) which is used
in this paper:
1. Use maximum likelihood to obtain estimates of β and σ for the regression of the
inverse of the eco-eﬃciency scores on the environmental variables using (4).
2. Loop over the following steps L times to obtain a set of bootstrap estimates for the
parameters β and σ :
B1 = (β∗
, σ∗
)b
L
b=1
(5)
10
(a) For each observation draw i from the N(0,σ ) distribution with left-truncation
at (1- ziβ).
(b) Compute for each observation Σ∗
i = ziβ + i
(c) Deﬁne p∗
i = pi and v∗
i = vi
Σi
Σi
for each observation.
(d) Compute the DEA eco-eﬃciency scores in (3) again by replacing vi and pi with
v∗
i and p∗
i and take the inverse of the eco-eﬃciency scores.
3. Calculate for each observation the bias-corrected estimator deﬁned as Σi = Σi −
BIAS(Σi) using the original inverse of the eco-eﬃciency and the bootstrapped
estimates obtained in Step 2.
4. Use maximum likelihood again to estimate the truncated regression, but using Σi
as the dependent variable. Obtain estimates of β and σ
5. Apply the following steps L times to obtain a set of bootstrap estimates:
B2 = (β∗
, σ∗
)b
L
b=1
(6)
(a) For each observation draw i from the N(0,σ ) distribution with left-truncation
at (1- ziβ).
(b) As done in Step 2.b , compute for each observation Σ∗∗
i = ziβ + i.
(c) Use maximum likelihood to estimate the truncated regression with Σ∗∗
i as dependent
variable and zi as explanatory variable. Obtain estimates of β∗ and
σ∗
6. Finally, construct estimated conﬁdence intervals for β and σ by using the original
estimates of β and σ and the bootstrap values obtained in B2.
11
3. Data description
The data set comes from a survey aimed at analysing the environmental performance
of the dairy sector in Asturias, Spain. A questionnaire was speciﬁcally designed for
the purpose of obtaining information on individual pollutants which could then be
aggregated using standard conversion factors into a series of environmental pressures.
Aside from pollutants, a novel aspect of the questionnaire is that it covered information
on the attitudes of farmers towards aspects of environmental management and some
socio-economic characteristics. The data collected correspond to the year 2010.
A total of 59 farmers responded to the questionnaire. The farms which took part
in the survey all belong to one of two groups of a Dairy Cattle Management Program
developed by the Board of Agriculture and Fisheries of the Principality of Asturias,
Spain. These management groups are formed by a minimum of 25 dairy farms and
each group is assigned a technical specialist who is responsible for collecting and processing
technical and economic data from each farm and who also provide advice on
how to improve the management of the farms. This information is then used to produce
monthly and annual reports on the performance of the farms. We combine this
economic information with the environmental data from the questionnaire to construct
our dataset. Given that there were missing values for some of the variables we wished
to consider, the ﬁnal sample comprises 50 farms.
To calculate eco-eﬃciency scores, we need information on economic value added and
environmental damage. Economic value added (Econvalue) is deﬁned as the diﬀerence
between revenues from milk production and direct costs, where revenues from milk
production comprise milk sales and the value of in-farm consumption of milk. Regarding
environmental damage, we use information on nutrients balances and greenhouse gas
12
emissions. The nutrients balances, which measure the extent to which a farm is releasing
nutrients into the environment, are deﬁned as the diﬀerence between the inﬂows and
outﬂows of nitrogen (SurplusN ), phosphorus (SurplusP) and potassium (SurplusK), all
measured in total kilograms. These environmental pressures are constructed using the
farm gate balance approach and are calculated as the diﬀerence between the nutrient
content of farm inputs (purchase of forage, concentrates, mineral fertilizers and animals,
legume ﬁxation of nitrogen in the soil and atmospheric deposition) and the nutrient
content of outputs from the farm (milk sales and animal sales).
The volume of greenhouse gas emissions represents an indicator of the contribution
of the farm to global warming. The dataset contains information on the emissions of
carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) coming both from
direct and indirect emissions. As each greenhouse gas has a diﬀerent global warming
potential, we convert each greenhouse gas into CO2 equivalents to make them comparable.
The variable used to denote greenhouse gas emissions is therefore measured as
thousands of kilos of carbon dioxide released into the atmosphere (CO2).
As determinants of eco-eﬃciency, we have information on a series of socio-economic
characteristics and attitudes of farmers. The socio-economic variables we choose are: the
age of the farmer (Age); the number of hours of speciﬁc agricultural training that the
farmer carried out during the year of the sample (Training); and a variable capturing
expected future prospects which is deﬁned as a dummy variable taking the value 1
if the farmer considered that the farm would continue to be in operation ﬁve years
later (Prospects). A priori, we would expect eco-eﬃciency to be negatively related to
age (i.e., young farmers should be more eco-eﬃcient) and positively related to speciﬁc
professional training and the expectation that the farm continue.
13
Three attitudinal variables are constructed from information gathered on farmers’
attitudes towards the management of nutrients and greenhouse gas emissions and their
attitudes towards environmental regulation. On a ﬁve-point Likert scale, respondents
had to state whether they strongly disagree (1), disagree (2), neither agree nor disagree
(3), agree (4) or strongly agree (5), with a series of statements regarding their
habits and attitudes towards aspects of environmental management. Thus, the variables
HabitsCO2 and HabitsNutrients measure farmers’ attitudes towards the importance
of measuring and managing greenhouse gas emissions and the nutrients balance respectively.
These are constructed as dummy variables that take the value 1 if respondents
stated that they agreed or strongly agreed that measurement and management of these
pressures was important, and 0 otherwise. The ﬁnal variable is a binary indicator measuring
attitudes towards environmental regulation (Regulation). Here, farmers were
asked whether they believed that environmental regulation was necessary. The variable
is also constructed such that it takes the value 1 if respondents stated that they
agreed or strongly agreed that environmental regulation should be tightened, i.e., that
the farmer believes that existing regulation is insuﬃcient, and 0 otherwise.
Some descriptive statistics of the variables used for measuring eco-eﬃciency and
the determinants of estimated eco-eﬃciency are presented in Table 1. [INSERT TABLE 1 ABOUT
HERE]
4. Results
4.1. Measuring eco-eﬃciency
Some summary statistics of the eco-eﬃciency scores estimated using equation (3) are
presented in Table 2, and a graph illustrating theindividuals scores is presented in Figure
2. These eco-eﬃciency scores represent the potential equi-proportional (or radial)
14
reduction of the four environmental pressures considered while maintaining economic
value added constant. The mean eco-eﬃciency score for the dairy farms in our sample
was 0.632, implying that the dairy farmers could reduce their environmental pressures
by an average of 36.8% while maintaining their economic value added. Five farms had
an estimated score of 1 and were thus found to be eco-eﬃcient, while the most ecoineﬃcient
farm had a score of 0.232, implying that it could maintain its level of value
added while reducing pressures by almost 87%. [INSERT TABLE 2 ABOUT
HERE]
[INSERT FIGURE 2 ABOUT
HERE]
Using the descriptive statistics for the environmental pressures in Table 1, the estimated
eco-eﬃciency scores imply that the representative farm could reduce it nitrogen
surplus by 0.368 x 5,966 kg = 2,194 kg, with corresponding reductions in phosphorus
and potassium surpluses of 1,019 kg and 771 kg respectively. Greenhouse gas emissions,
in turn, could be reduced by 157,161 kg of CO2 equivalent.
Our results point to a high level of eco-ineﬃciency in Spanish dairy farms, but the
fact there are no previous studies of eco-eﬃciency of dairy farms using the methodology
we use precludes a direct comparisons of our results with existing results in the
literature. However, Picazo-Tadeo et al. (2011) and Picazo-Tadeo et al. (2012) have
used this methodology to assess eco-eﬃciency of Spanish farmers from other sectors.
Picazo-Tadeo et al. (2011) estimate eco-eﬃciency for a sample of 171 crop farmers in
the Castille-Leon region and found a potential average equi-proportional reduction of
environmental pressures of 44%, whereas in a study of 55 olive farmers in southern
Spain, Picazo-Tadeo et al. (2012) ﬁnd that environmental pressures could be equiproportionally
reduced by 46%. For dairy farms, the closest study to ours is perhaps
that of Iribarren et al. (2011), which uses data from the neighbouring Spanish region
of Galicia. Using a sample of 72 dairy farms, these authors combine Life Cycle As-
15
sessment (LCA) and DEA to estimate eco-eﬃciency and ﬁnd potential reduction of
environmental pressure ranging from 23% to 31%.
4.2. Explaining eco-eﬃciency
The estimated coeﬃcients and conﬁdence intervals from the truncated regression are
presented in Table 3.
[INSERT TABLE 3 ABOUT
HERE]The model we use performs quite well in the sense that all the explanatory variables
were found to be highly statistically signiﬁcant. The variables chosen therefore
permit some insights into the factors inﬂuencing eco-eﬃcient outcomes. Given that
the dependent variable is deﬁned as the inverse of the eco-eﬃciency scores, a positive
estimated coeﬃcient indicates greater eco-ineﬃciency and hence lower eco-eﬃciency,
whereas a negative estimated coeﬃcient indicates lower eco-ineﬃciency and therefore
a higher level of eco-eﬃciency.
With this in mind, and beginning with the socio-economic variables, the positive
estimated coeﬃcient on the age variable indicates that older farmers are less ecoineﬃcient
so that the age of the farmer has a negative impact on eco-eﬃciency. This
result is in line with Reinhard et al. (2002) who found that younger farmers are more
likely to be more knowledgeable about environmentally-friendly technological progress.
The estimated coeﬃcient for the variable capturing the future prospects of the farm
is negative, implying that farmers that will continue their activity in the next ﬁve years
are running their farms more eco-eﬃciently. A similar result was found by Van Passel
et al. (2009) and suggests that farmers that envisage their farm continuing - either run
by themselves or by their successors - operate in a relatively more sustainable way.
16
The agricultural training variable had a negative coeﬃcient, indicating that the
greater the number of hours of speciﬁc agricultural instruction, the higher the level of
eco-eﬃciency. However, it should be noted that this eﬀect was found to be relatively
weak at as the coeﬃcient has quite a small value. In a similar vein, Picazo-Tadeo
et al. (2011) included training as an explanatory variable of eco-eﬃciency scores for
crop farmers and found it had no eﬀect. This would seem to suggest that agricultural
training programs could be improved in order to become a more eﬀective tool in the
eﬀort to reduce environmental damages.
As might be expected, self-reported positive habits regarding nutrients balance
management and emissions management improve eco-eﬃciency. This implies that farmers’
responses to questions about the importance they attribute to management and
control of environmental pressures provide, as we would hope, an accurate reﬂection of
their actual behaviour, as manifested by their higher levels of eco-eﬃciency. We take
this as lending credibility to the responses in the survey and hence to the data used.
Finally, the positive coeﬃcient for the variable Regulation indicates that farmers
that more strongly favour the need for regulation are less eco-eﬃcient. In other words,
the more eco-eﬃcient farmers see less need for regulation and see no need to make it
stricter.
5. Conclusions
Assessments of eco-eﬃciency can shed light on the environmental performance of agricultural
producers and provide a potentially valuable source of information for policymakers
in a context where sustainability is an ever-growing concern. We have estimated
eco-ineﬃciency scores for a sample of dairy farms in northern Spain and found that
17
they are highly eco-ineﬃcient. On average, these farmers could maintain their present
economic value added while reducing the environmental damage associated with their
productive activity measured by a series of environmental pressure indicators. We ﬁnd
that pressures could be reduced by over 37%.
We investigate the impact of socio-economic characteristics and farmers’ attitudes
towards the environment on their eco-eﬃciency. We ﬁnd that younger farmers are
more eco-eﬃcient, as are farmers who envisage their operation continuing for at least
the following ﬁve years. Participation in speciﬁc agricultural training schemes had a
positive but weak eﬀect on eco-eﬃciency, whereas farmers who believe environmental
regulation should be strenghtened were more eco-ineﬃcient. We also found a positive
relationship between the self-reported environmental habits of farmers and their actual
environmental performance as reﬂected in higher estimated eco-eﬃciency levels.
The main implication of our study is that there is substantial room for costless
-in terms of economic value added - improvements in the environmental performance
of dairy farmers. Better management of resources and possibly improved technical
eﬃciency would lead to relatively large reductions in nutrients balances and greenhouse
gas emissions.
Our results also imply that policymakers wishing to reduce environmental damage
should encourage young farmers to become more involved in the running of dairy farms.
Moreover, training schemes should be improved in order to strengthen their positive
eﬀect on environmental performance. The beneﬁts of environmental regulation should
be clearly explained and measures to reduce the cost to farmers of implementing such
regulation should be explored. Finally, the responses of the farmers in our sample with
regard to the importance they attribute to management and control of environmental
pressures are consistent with observed eco-eﬃcient behaviour. This implies that survey
18
information on farmer attitudes may be a valuable source of information on actual
environmental performance.
Notes
1Picazo-Tadeo et al. (2011),while lacking the data to do so themselves, stressed the importance
of taking into account attitudinal variables in the analysis as, in their own words,
“other farmer features, including psychological aspects such as environmental concerns, should
be considered in the future to explain eco-eﬃciency”.
2The transformed eco-eﬃciency measure ranges from 1 to inﬁnity.
References
Ball, V. E., Lovell, C. A. K., Nehring, R. F. and Somwaru, A.: 1994, Incorporating
Undesirable Outputs into Models of Production, Cahiers D´Economie et Sociologie
Rurales 31, 60–73.
Basset-Mens, C., Ledgard, S. and Boyes, M.: 2009, Eco-eﬃciency of intensiﬁcation
scenarios for milk production in new zealand, Ecological Economics 68, 1615–1625.
Callens, I. and Tyteca, D.: 1999, Towards indicators of sustainable development for
ﬁrms: A productive eﬃciency perspective, Ecological Economics 28, 41–53.
CEAS: 2000, The environmental impact of dairy production in the EU: Practical options
for the improvement of the environmental impact, Technical report, Report
submitted to European Commission (DGXI ).
Coelli, T., Prasada Rao, D., O’Donnell, C. and Battese, G.: 2005, An Introduction to
Eﬃciency and Productivity Analysis, 2nd edition edn, Springer, New York.
19
COWI: 2001, Cleaner production assessment in dairy processing, Technical report, Report
prepared for United Nations Environment Programme and Danish Environmental
Protection Agency .
Defrancesco, E., Gatto, P., Runge, F. and Trestini, S.: 2008, Factors aﬀecting farmers’
participation in agri-environmental measures: A northern italian perspective, Journal
of Agricultural Economics 59, 114–131.
FAO and IDF: 2011, Guide to good dairy farming practice. animal production and
health guidelines, Technical Report 8, Rome.
F¨are, R., Grosskopf, S. and Lovell, C.: 1985, The Measurements of Eﬃciency of Production,
Kluwer- Nijhoﬀ, Boston.
F¨are, R., Grosskopf, S., Lovell, C. and Pasurka, C.: 1989, Multilateral productivity
comparisons when some outputs are undesirable: A nonparametric approach, Review
of Economics and Statistics 71, 90–98.
Fried, H., C. L. and Schmidt, S.: 2008, The Measurement of Productive Eﬃciency and
Productivity Growth, Oxford University Press, Oxford.
Iribarren, D., Hospido, A., Moreira, M. and Feijoo, G.: 2011, Benchmarking environmental
and operational parameters through eco-eﬃciency criteria for dairy farms,
Science of the Total Environment 409, 1786–1798.
Kortelainen, M.: 2007, Dynamic environmental performance analysis: A Malmquist
index approach, Ecological Economics 64, 701–715.
Kuosmanen, T. and Kortelainen, M.: 2005, Measuring eco-eﬃciency of production with
data envelopment analysis, Journal of Industrial Ecology 9(4), 59–72.
Latruﬀe, L., Davidova, S. and Balcombe, K.: 2008, Application of a double bootstrap
to investigation of determinants of technical eﬃciency of farms in central europe,
Journal of Productivity Analysis 29(2), 183–191.
20
Lauwers, L.: 2009, Justifying the incorporation of the materials balance principle into
frontier-based eco-eﬃciency models, Ecological Economics 68, 1605–1614.
OECD: 1998, Eco-Eﬃciency, Organization for Economic Co-operation and Development,
OECD, Paris.
OECD: 2004, Agriculture, trade and the environment: The dairy sector, Technical
report, OECD Publishing, Paris.
Oenema, O. and Pietrzak, S.: 2002, Nutrient Management in Food Production: Achieving
Agronomic and Environmental Targets, AMBIO: A Journal of the Human Environment
31(2), 159–168.
Olson, K. and Vu, L.: 2009, Economic eﬃciency in farm households: trends, explanatory
factors, and estimation methods, Agricultural Economics 40(5), 587–599.
Oude Lansink, A. and Bezlepkin, I.: 2003, The Eﬀect of Heating Technologies on CO2
and Energy Eﬃciency of Dutch Greenhouse ﬁrms, Journal of Environmental Management
68, 73–82.
Oude Lansink, A. and Silva, E.: 2003, CO2 and Energy Eﬃciency of Diﬀerent Heating
Technologies in the Dutch Glasshouse Industry, Environmental and Resource
Economics 24, 395–407.
Picazo-Tadeo, A. J., Beltran-Esteve, M. and Gomez-Limon, J.: 2012, Assessing ecoeﬃciency
with directional distance functions, European Journal of Operational Research
220, 798–809.
Picazo-Tadeo, A., Reig-Mart´ınez, E. and G´omez-Lim´on, J.: 2011, Assessing farming
eco-eﬃciency: a data envelopment analysis approach, Journal of Environmental Management
92(4), 1154–1164.
Pittman, R. W.: 1983, Multilateral Productivity Comparisons with Undesirable Outputs,
Economic Journal 93, 883–891.
21
Reinhard, S., Lovell, C. A. K. and Thijssen, G. J.: 2002, Analysis of environmental
eﬃciency variation, American Journal of Agricultural Economics 84(4), 1054–1065.
Reinhard, S., Lovell, C. and Thijssen, G.: 1999, Econometric estimation of technical
and environmental eﬃciency: an application to Dutch dairy farms, American Journal
of Agricultural Economics 81, 44–60.
Schmidheiney, S.: 1993, Changing course: A global business perspective on development
and the environment, Technical report, MIT Press, Cambridge, MA.
Simar, L. and Wilson, P.: 2007, Estimation and inference in two-stage, semi-parametric
models of production processes, Journal of Econometrics 136, 31–64.
Tyteca, D.: 1996, On the measurement of the environmental performance of ﬁrms - a
literature review and a productive eﬃciency perspective, Journal of Environmental
Management 68, 281–308.
Tyteca, D.: 1997, Linear Programming Models for the Measurement of Environmental
Performance of Firms: Concepts and Empirical results, Journal of Productivity
Analysis 8, 183–197.
Tyteca, D.: 1999, Sustainability indicators at the ﬁrm level, Journal of Industrial Ecology
2, 61–77.
Van Passel, S., Van Huylenbroeck, G., Lauwers, L. and Mathijs, E.: 2009, Sustainable
value assessment of farms using frontier eﬃciency benchmarks, Journal of Environmental
Management 90, 3057–3069.
Weninger, Q.: 2008, Economic beneﬁts of management reform in the gulf of mexico
grouper ﬁshery: A semi-parametric analysis, Environmental and Resource Economics
41(4), 479–497.
Wynn, G., Crabtree, B. and Potts, J.: 2001, Modelling farmer entry into the environmentally
sensitive area schemes in scotland, Journal of Agricultural Economics
22
52, 65–82.
	  
Environmental
Pressure
p1
●!A!
T!
O!
E!
●!
Eco-efficient frontier
Environmental
Pressure
P2
Fig. 1: Pressure-generating technology set and eco-eﬃcient frontier
23
0.00	  
0.10	  
0.20	  
0.30	  
0.40	  
0.50	  
0.60	  
0.70	  
0.80	  
0.90	  
1.00	  
1	   2	   3	   4	   5	   6	   7	   8	   9	  10	  11	  12	  13	  14	  15	  16	  17	  18	  19	  20	  21	  22	  23	  24	  25	  26	  27	  28	  29	  30	  31	  32	  33	  34	  35	  36	  37	  38	  39	  40	  41	  42	  43	  44	  45	  46	  47	  48	  49	  50	  
Eco-­‐ﬃciency	  scores	  
Farm	  
Eco-­‐eﬃciency	  scores	  of	  sample	  of	  Spanish	  dairy	  farms:	  2010	  
Fig. 2: Individual farm eco-eﬃciency scores
Table 1: Descriptive Statistics- Ecoeﬃciency dairy farms in Asturias
Variable Mean Std. Dev Min Max
Econvalue 77137 40423 16959 168848
Environmental Pressures
SurplusN 5966 4705 0 16244
SurplusP 2770 2168 0 7714
SurplusK 2096 1681 0 5975
CO2 427.3 142.0 207.2 828.6
Eco-eﬃciency Determinants
HabitsCO2 0.09 0.29 0 1
HabitsNutrients 0.77 0.43 0 1
Age 45.98 7.97 29 65
Prospects 0.98 0.14 0 1
Regulation 0.58 0.50 0 1
Training 45.14 63.10 0 400
Table 2: Descriptive statistics of computed Eco-eﬃciency scores
Mean Std. Dev Min 1st Qt Median 3rd Qt Max
Radial
0.6322 0.2319 0.2039 0.4772 0.6158 0.8342 1
eco-eﬃciency
24
Table 3: Truncated regression and bootstrapped conﬁdence intervals
(95% conﬁdence) (90% conﬁdence)
Variables Estimated Lower Upper Lower Upper
parameter bound bound bound bound
HabitsCO2 -0.6891 -0.6897 -0.2879 -0.6606 -0.3179
HabitsNutrients -0.2307 -0.3940 -0.1680 -0.3815 -0.1850
Age 0.0082 0.0013 0.0144 0.0025 0.0134
Prospects -2.1438 -2.4408 -1.8411 -2.4010 -1.8781
Regulation 0.2302 0.1896 0.4041 0.2046 0.3919
Training -0.0019 -0.0031 -0.0009 -0.0028 -0.0010
Sigma 0.1611 0.1097 0.1746 0.1147 0.1703