Eo

Aa

b

a

ARRA

KODEOMD

I

ore

c

va

0h

Land Use Policy 36 (2014) 275– 284

Contents lists available at ScienceDirect

Land Use Policy

jou rn al hom epage: www.elsev ier .com/ locate / landusepol

nvironmental efficiency and the impact of regulation in drylandrganic vine production

na M. Aldanondo-Ochoaa,∗, Valero L. Casasnovas-Olivaa, Amaia Arandia-Miurab

Universidad Pública de Navarra, Departamento de Gestión de Empresas, Pamplona 31006, SpainGobierno de Navarra, Servicio de Innovación y Transferencia de Tecnología, Pamplona, Spain

r t i c l e i n f o

rticle history:eceived 21 July 2012eceived in revised form 7 August 2013ccepted 10 August 2013

eywords:rganic agricultureryland farmingnvironmental efficiencyrganic regulationeta-production frontierEA

a b s t r a c t

Organic agriculture figures prominently in the policies adopted by the EU to improve the environ-mental impact of agriculture. It may also potentially provide other benefits such as high-quality,health-enhancing food products and advancements in rural development. Recent years have broughtnew research to assess the environmental and economic implications of organic conversion. Economicefficiency comparisons between organic and conventional farms have been extended to include environ-mental performance. The inclusion of this variable in efficiency analysis may be useful when assessing thepotential impact of suggestions to improve environmental regulations and policies. This paper applies theenvironmental efficiency model to the analysis of different technologies and calculates productivity andefficiency with and without environmental impacts. In the empirical part of the paper Data EnvelopmentAnalysis (DEA) and bootstrap techniques are applied to detect and measure differences between organicand conventional agriculture aggregate efficiency and productivity in a sample of vineyard farms operat-ing in semiarid, non-irrigated conditions in Navarre (Spain), taking farms’ nitrogen surplus and pesticidetoxicity indicators to consideration. The results for these particular agronomic conditions suggest thatorganic agriculture is more environmentally efficient than conventional agriculture in dryland farming,in that it achieves a more favorable production to environmental impact ratio. Nevertheless, conversionto organic production methods for extensive vine cultivation under arid conditions does not guaranteesubstantial environmental gains, since the organic farms in our sample do not display inferior levels ofpollution emissions per unit input as extensive conventional production. The overall environmental effi-ciency of organic farming is largely attributable to the fact that organic farms come closer to the frontierof their own technology. We find no significant technological differences in environmental productivity,however. In terms of policy implications, these findings suggest that the tightening of specific envi-

ronmental restrictions in organic standards should involve consideration of technological differences inenvironmental productivity between organic and other alternative technologies. If organic technology isless productive, more restrictive regulation could undermine the economic viability of farms, and thusundermine the other benefits of organic farming. The results also indicate that, at the local level, it couldbe convenient to address part of organic subsidies to further improvements in the control of pollution

cides.

from fertilizers and pesti

ntroduction

Increasing public concern for the environmental externalitiesf agricultural production has given organic farming an importantole in policies aimed at improving the impact of agriculture on the

nvironment in the European Union.

From the agri-environmental policy perspective, organic agri-ulture may be considered a voluntary technological standard

∗ Corresponding author. Tel.: +34 948 169 633; fax: +34 948 169 404.E-mail addresses: alda@unavarra.es (A.M. Aldanondo-Ochoa),

alero.casasnovas@unavarra.es (V.L. Casasnovas-Oliva),maia.arandia.miura@navarra.es (A. Arandia-Miura).

264-8377/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.landusepol.2013.08.010

© 2013 Elsevier Ltd. All rights reserved.

(OECD, 2010). Organic farming standards (Council Regulation (EC)No 834/2007) prohibit the use of synthetic chemical fertilizersand pesticides, control the use of certain inputs and specify aseries of required agricultural practices. Farmers certifying com-pliance with these standards receive a subsidy per unit of areafarmed in exchange for their contribution to public environmen-

tal conservation (Offermann et al., 2009; Sauer and Park, 2009).1 Afarmer’s compliance with organic farming standards is judged notin terms of environmental performance but in terms of the farmer’s

1 Subsidies to organic farmers under the EU agri-environmental policy since 1993,have been included in rural development programs and, may increase with theintroduction of new grants for organic producers in 2013.

2 Land U

cpe

oooiotetpcotPsaac2e“cte(cooasecrsttr

oocaeaaapacetoiTiaa2

a

output efficiency across industries using individual firm-leveldata. Individual efficiency scores are computed at an early stage bymeans of Data Envelopment Analysis (DEA). We chose this method

3 Despite conceptual differences, the terms “environmental externality”, “envi-ronmental impact”, “pollutant emissions”, and “environmental pressurewill be usedindistinctly throughout this paper, the main point being that agricultural productionhas an impact on the public environment. For a full clarification of the meanings ofthese terms, see OECD (2010).

4 According to Färe et al. (1989) firms have the same technology: the envi-ronmentally regulated technology (weak disposability) and the environmentallynon-regulated technology (strong disposability). However, they are subject to dif-ferent levels of environmental regulation given by the current cap on pollutantemissions (per unit input). Regulation on organic agriculture is basically techno-logically oriented. Only rarely does it restrict the quantity of pollutant emissions,

76 A.M. Aldanondo-Ochoa et al. /

hoice of inputs and technology, which may affect farm economicerformance. This raises two major issues potentially affecting EUnvironmental policy.

The first issue involves the environmental effectiveness ofrganic farming regulations. Specifically, the question is whetherrganic standards suffice to reduce the environmental pressuref agriculture or further restrictions are required. This matter ismportant because both the market and policy-makers considerrganic agriculture a more environmentally-friendly food produc-ion system. The effectiveness of organic regulation for generatingnvironmental benefits depends on the environmental potential ofhe organic production system and on how effectively agriculturalractices are applied. As far as environmental potential is con-erned, various studies based on experimental research show thatrganic farming may play a key role in environmental conserva-ion and natural resource management (Korsaeth and Eltun, 2000;imentel et al., 2005; Thomassen et al., 2008). Additionally, varioustudies have shown that when it comes to preserving soil quality,nd protecting surface water, climate, air and biodiversity, organicgriculture generally performs better than other systems whenompared in terms of units of area farmed (Casey and Holden,006; Cederberg and Mattsson, 2000; Haas et al., 2001; Holet al., 2005; Rahmann, 2011; Stolze et al., 2000). Nevertheless, theconventionalization” hypothesis sustains that organic farmingarries the risk of becoming more intensive and industrialized,hereby devaluing its role as a sustainable alternative (Buckt al., 1991; Reed, 2005), and reducing its environmental benefitsDarnhofer et al., 2010). Mansfield (2004) suggests that thisonventionalization process may be due to the institutionalizationf organic farming, which creates a gap between the complexityf the values and principles underlying the organic movementnd the necessary simplification of regulatory measures into aeries of prohibited and permitted production practices.2 Padelt al. (2009) suggest that the environmental concerns raised by theonventionalization hypothesis could be addressed by introducingestrictions on the use of organic and non-organic inputs, andetting explicit objectives aimed at achieving a production systemhat is more in keeping with environmental quality. This appearso be the direction being taken by the European Union in itsegulatory reform program (Padel et al., 2009).

The second issue, which concerns the environmental efficiencyf organic farming, arises from uncertainty as to the capacity ofrganic farming to achieve environmental objectives at a lowerost than is possible with alternative production systems. Therere two important sides to this question. Schader (2009), forxample, considers that, with a limited budget, it is vital to selectgri-environmental policies that can achieve the same purposet a lower cost. Zimmermann et al. (2011) note, in addition, thatgriculture plays various social roles, including not only foodroduction but also the continuation of economic activity in ruralreas: hence the importance of checking production system effi-iency. Is the maximum output being achieved with the minimumnvironmental impact? In this respect, various studies have shownhat the estimated environmental advantage of organic versusther agricultural production systems diminishes if the compar-son is made in terms of environmental impact per unit output.he results of which show it to be superior in some environmentalndicators, such as green house emissions, energy consumptionnd acidification and inferior in others, such as nitrate leaching

nd eutrophication (Backer et al., 2009; Cederberg and Mattsson,000; De Boer, 2003; Tuomisto et al., 2012).

2 Guthman (2004) also considers the influence of agribusiness on the setting ofgricultural norms and practices in California.

se Policy 36 (2014) 275– 284

These two issues – the environmental effectiveness oforganic regulations and the environmental efficiency of organicagriculture at farm level – provide the focus of this paper. We adoptthe approach to firm-level environmental performance evaluationused in the literature on productive efficiency analysis (Dyckhoffand Allen, 2001; Färe et al., 1989; Hailu and Veeman, 2001). Thisentails comparison of organic and conventional farm performanceusing multiple environmental indicators3 to measure differences inoutput per unit of environmental impact and input (environmentalefficiency) and differences in environmental impact per unit input(environmental effectiveness).

With respect to methodological issues, it is worth notingthat previous studies (Arandia and Aldanondo-Ochoa, 2008;Kumbhakar et al., 2009; Oude-Lansink et al., 2002) consider organicand conventional agriculture different technologies. Organic farmsalso support more restrictive environmental regulations. Therefore,when comparing the efficiency of organic and conventional agricul-ture it is necessary to develop an analytical framework allowingfor comparisons between firms that are potentially disparate interms of the technology they use or the stringency of environ-mental regulation to which they are subject. The inclusion of themeta-production function (Hayami and Ruttan, 1971) in efficiencyanalysis provides a basis for the evaluation of firms in different tech-nology groups (Arandia and Aldanondo-Ochoa, 2008; Battese et al.,2004; Kumbhakar et al., 2009; O’Donnell et al., 2008; Oude-Lansinket al., 2002). It allows a decomposition of output in technologicalproductivity differences (technology gap) and efficiency of firmsof a group relative to the best practice in this group. At the sametime, differences in output for firms under different environmen-tal regulation4 have been attributed to environmental technicalefficiency and to the impact of environmental regulation on pro-ductivity (Färe et al., 1989). The impact of environmental regulationis the reduction in output forfeited by farms by using technologiesthat reduce pollutant emissions. The greater the amount forfeited,the more restrictive the regulation is considered. Then, the impactof regulation on productivity gives an indirect proxy for the regula-tory effectiveness.5 In this paper, we combine these two approachesto propose a decomposition of output into an index of environmen-tal technical efficiency, an index of environmental productivity andan index of the impact of regulation on productivity.

As well as a workable analytical framework, it is also importantto select a reliable procedure to measure and compare between-group efficiency. Recently, Simar and Zelenyuk (2007) proposedbalanced subsampling bootstrap techniques to compare aggregate

which is what actually determines the value of the environmental effectivenessindex. When different groups of firms employ different types of technology, as inthe case in hand, the only way to estimate the environmental regulation impactindex proposed by Färe et al. (1989) it by constructing a single production set andmeasuring all firms against the same technological efficiency frontier, which in thiscase is the meta-frontier of efficiency.

5 Given the difficulty of aggregating physical quantities of different environmen-tal impacts, we use the impact of regulation index as proxy for the environmentalimpact per unit input.

Land Use Policy 36 (2014) 275– 284 277

tfs

ccgouoMf

dpt

M

emoitt(fr

I

(KOaafptico

unWauridoaTm

I

m

oPKd

A.M. Aldanondo-Ochoa et al. /

o compare the performance of organic and conventional farmingor its explanatory power when evaluating the two productionystems.

The empirical section of this paper compares the performance ofonventional versus organic farms engaged in extensive vineyardultivation in the Navarre region of Spain, taking into account nitro-en surplus and the potential toxicity of pesticides. The interestf this case lies in the semiarid, non-irrigated farming conditionsnder which all the sample farms are managed. This is the typef production system typically found in extensive farms in theediterranean and other arid areas. The study is based on findings

rom survey of organic and conventional farms conducted in 2004.The paper is organized as follows. In the next section, we

escribe the methodology. Section ‘Data and sample description’resents the data set, and the following sections contain the results,he discussion and the conclusions of the study.

ethodology

In this section, we begin by decomposing the overall technicalfficiency (without considering environmental impact), into threeeasures: an index of environmental technical efficiency, an index

f (technological differences) environmental productivity and anndex of the impact of regulation on productivity. We then specifyhe DEA (Data Envelopment Analysis) programs used to estimatehem. The aggregate efficiency indices for the two groups of farmsconventional and organic) and the bootstrap method used to per-orm the sensitivity analysis are presented in Appendices B and Cespectively.

ndividual DEA efficiency and productivity indices

We begin by applying the meta-production function approachArandia and Aldanondo-Ochoa, 2008; Battese et al., 2004;umbhakar et al., 2009; Mayen et al., 2010; O’Donnell et al., 2008;ude-Lansink et al., 2002) to measure differences in the over-ll technical environmental efficiency of organic and conventionalgriculture relative to the meta-frontier of efficiency. These dif-erences can arise either from the environmental technologicalroductivity (which refer to the potential of either technologyo produce a higher output with less inputs and environmentalmpact) or from differences in farm environmental technical effi-iency relative to the farm’s own group frontier (conventional orrganic), or from both (Kumbhakar et al., 2009).

To define the environmental technical efficiency indices, wese the same specification of environmental technology or tech-ology with undesirable outputs as in Hailu and Veeman (2001).e assume that the farmer uses M inputs to produce N desir-

ble outputs and R undesirable outputs or pollutant emissions. Thendesirable outputs are non-freely disposable (the technology isegulated): to prevent them, farmers would need to increase theirnputs or reduce the desirable outputs. Let x ∈ R+

M , y ∈ R+N , b ∈ R+

Renote a vector of inputs, a vector of desirable outputs and a vectorf undesirable outputs, respectively. The technology consisting ofll feasible (y, b, x) is denoted by T = {(y, b, x): x can produce (y, b)}.his technology satisfies, among other conditions,6 the following

onotonicity condition.7

f (y, b, x) ∈ T and y ≥ y′, b′ ≥ b and x′ ≥ x, then (y′, b′, x′) ∈ T.

6 This production set is closed and convex and satisfies the weak axiom of profitaximization.7 This type of monotonicity specification reduces the bias in the desirable

utput efficiency score (Arandia and Aldanondo-Ochoa, 2011; Picazo-Tadeo andrior, 2009), allows substitution between undesirable outputs (Kortelainen anduosmanen, 2007; Kuosmanen and Kortelainen, 2005) and is analogous to weakisposability in the directional sense defined by Färe et al. (2006).

Fig. 1. Output-oriented environmental technical efficiency with different, technolo-gies.

Under monotonicity, the output set P(x) = {(y, b): (y, b, x) ∈ T}satisfies the following condition:

If (Y, b) ∈ P(x) and y ≥ y′, b′ ≥ b, then (y′, b′) ∈ P(x). (1)

Fig. 1 depicts the output sets (y, b) for organic and conventionaltechnology. Each dot represents the combination of one output, y,and a pollution emission, b, produced by each farm. In this figure,conventional farms are indicated by the letter “C” and organic farmsare indicated by the letter “R”. Under condition (1), the organicand conventional agricultural output sets will be bounded by thelines hR1 and mC1, respectively (Färe and Grosskopf, 2003). Theline C1-C2-C3 is the technical environmental efficiency frontierof the conventional production system and R1-R2-R3-R4 that ofthe organic production system. The efficient farms produce moredesirable outputs and/or fewer pollutant emissions than inefficientfarms using the same amount of factors.

The overall technical environmental efficiency frontier or meta-production frontier, represented on the graph by R1-R2-R3-C3, isthe envelope of the organic and conventional efficiency frontiers.This is the efficiency frontier for the entire sample of farms, bothconventional and organic, including environmental impacts.

Now let us consider an inefficient conventional farm, C. Webegin by defining two basic environmental efficiency measures forthis farm, C:

(1) The technical environmental efficiency for unit C relative tothe frontier of conventional farms is given by the quotientEffse = NC/NE, which measures the proportionate increase inoutput (keeping pollutant emissions and inputs constant) thatunit C must achieve in order to reach best practice for its ownproduction set.

(2) The overall technical environmental efficiency index for unit Crelative to meta-frontier is the ratio, Effsg = NC/NS. Effsg meas-ures the proportionate increase in output required by a givenfarm to reach the meta-production frontier, keeping pollutantemissions and inputs constant. This index shows which of thetwo groups of farms is environmentally more efficient: that is,achieves more output from the same quantity of inputs, while

producing the same amount of pollutant emissions.

The ratio between the two basic scores described above allowsus to identify the technology gap or the index of environmental

278 A.M. Aldanondo-Ochoa et al. / Land U

ptpswCvop

orFtsa(ba(

feltg(wti

erTia

ds

1kxmk), produces an amount of outputs (y1k, . . ., ynk) and emits anamount of pollutants (b1k, . . ., brk). The output efficiency measureEffs, aims to find the maximum potential yield increase that can be

Fig. 2. Meta-frontiers of efficiency and the impact of regulation.

roductivity, Prods = NE/NS, which measures the distance betweenhe efficiency frontier of each group and the metafrontier. It is com-uted from the quotient Effse/Effsg. If a farm’s current technology isuperior to other technologies, its Prods score will be 1. Otherwise, itill be less than 1. It should be noted that, for the conventional farm, organic technology is assumed to be more productive than con-entional technology. This score measures the difference in termsf environmental efficiency between organic and conventional bestractice.

For the second analysis, we use an approach similar8 to thatf Färe et al. (1989) to compare the stringency of environmentalegulation supported by organic versus conventional agriculture.or this we introduce the non-regulated technology and assumehat each farm can freely dispose of its undesirable outputs. Thisituation can be modeled “by assuming (that) all outputs — goods well as bad — are freely disposable, and then dropping the badones)” (Färe et al., 2007, p. 676). The resulting production set woulde Q(x) = {y: (y, x) ∈ T}, that is, the different combinations of desir-ble outputs that can be produced with a given quantity of inputsregardless of undesirable outputs).

In Fig. 2, R1-R2-R3-C3 represents the meta-production frontieror output set P(x) with undesirable outputs featured in Fig. 1 for thentire set of farms, both organic and conventional. The horizontaline, a-G-C3 is the meta-production frontier for the (unregulatedechnology) output set Q(x). Among all the farms shown on theraph, unit C3 stands out as achieving the highest level of output, yusing the same quantity of factors, x, see Appendix A). Thus, whene measure the overall technical efficiency of organic and conven-

ional farms, without considering environmental externalities, C3s the only efficient farm in the sample represented in Fig. 2.

We then define the overall technical efficiency index (Batteset al., 2004; Mayen et al., 2010; Oude-Lansink et al., 2002), of unit Celative to the meta- production frontier as the ratio, Effg = NC/NG.his index of overall technical efficiency assumes that, by increasing

ts output 1/Effg, while keeping its inputs x constant, unit C canttain the same output level as unit C3. This efficiency index takes

8 We describe our approach as similar because we adopt only Färe et al.’s (1989)ecomposition; not their definitions of weak and strong disposability, as have beeen before.

se Policy 36 (2014) 275– 284

a value of 1 when the unit is efficient and less than 1 when it isinefficient.

The ratio between the overall environmental technical effi-ciency index, Effsg, and the overall conventional technical efficiencyindex, Effg, gives us the impact-of-regulation index, Regula-tion = Effg/Effsg, which assumes that unit C can increase its output,Regulation = NS/NG, simply by increasing its pollutant emissions(per unit input). The Regulation index represents the output reduc-tion (per unit of input) to be sustained by a farm in order to controlits level of control pollutant emissions (per unit input). It is con-sidered that C must sustain a productivity loss in order to keeppollutant emissions per unit input at level N, which is assumed tobe determined by environmental regulations (Färe et al., 1989).9 Afarm with a level of pollutant emissions greater than or equal tothat of C3 is assumed to be under no environmental restrictionsand the Regulation index takes the value 1. For a farm with a levelof emissions lower than that of C3, Regulation is less than unity,because the farm is assumed to sacrifice some of its potential outputin order to control its pollutant emissions. The assessment of all theunits against the meta-frontiers of efficiency ranks them in orderof pollutant emissions per unit input. This provides an approxi-mate indication of environmental regulatory effectiveness. Scoreson this index also give some orientation regarding the environmen-tal farm-management practices.

Then, by means of these two procedures, the overall technicalefficiency index can be decomposed into:

Effg = Effsg ∗ Regulation = Effse ∗ Prods ∗ Regulation

This breakdown is extremely useful for examining both the envi-ronmental effectiveness and environmental efficiency of organicfarming. For example, higher Prods and Effse index scores and lowerRegulation index scores indicate, firstly, that organic farms are ableto produce higher output per environmental impact, and, further-more, that they are able to reduce their emissions per unit input,which would appear to suggest that organic farming is environ-mentally more efficient and that organic standards can be trulyenvironmentally effective.

DEA program for the estimation of individual Effs

DEA is a non-parametric efficiency estimation method in whichlinear programming is used to estimate the efficiency frontierand measure each unit’s distance from it. The DEA technical-environmental efficiency estimates obtained in this study indicatethe maximum potential output increase a farm can achieve withoutincreasing inputs or emissions. The DEA output efficiency modelused in this study treats pollution as an input (Kortelainen andKuosmanen, 2007; Kuosmanen and Kortelainen, 2005; Pittman,1983; Reinhard et al., 1999; Zhang et al., 2008). We use this modelto estimate an Effsg score for each farming unit in the full sample(conventional and organic farms) and then estimate the Effse foreach of the two subsamples.

Let us suppose a sample of k = 1, . . ., K farms. Each farm in samplek uses a quantity of inputs represented by the input vector (x , . . .,

9 We have mentioned already that Färe et al. (1989) define the concept for firmsusing the same technology. For farms using different technologies, it is possibleto construct a concept analogous to the impact of Regulation by taking the meta-efficiency frontiers as a reference. In conceptual terms, this would represent theproductivity reduction in order to maintain pollutant emissions per unit input atthe current level using the best technology available.

Land U

ab

tuiaec

vtetegmtcstumpcacthtr(t

D

E

AfipiAtpa

ingredient in pesticides on the health of farm workers, the healthof consumers, and biodiversity. The farms’ EIQF (EnvironmentalImpact Quotient of Farm) were calculated by aggregating the EIQ

11 Since DEA models are prone to be highly sensitive to measurement errors andoutliers, several outlier tests (Chen and Johnson, 2010; Simar, 2003; Wilson, 1993)were carried out. A thorough check of the sample resulted in 6% of the initial sample

A.M. Aldanondo-Ochoa et al. /

chieved with the same level of inputs and pollution. It is calculatedy means of the following linear programming problem:

Effsk = min ˇ

subject to∑

k

�kynk ≥ ynk/ n = 1, . . ., N

∑

k

�kbrk ≤ brk r = 1, . . ., R

∑

k

�kxmk ≤ xmk m = 1, . . ., M

�k ≥ 0 k = 1, . . ., K∑

k

�k = 1

(2)

This is a standard output-oriented DEA model, in which pollu-ion is treated as an input. The index is equal to (less than) 1 if thenit is efficient (inefficient). By solving this model for every farm

n the sample, we are able to estimate the potential proportion-te output increase 1/ˇ that it is able to achieve in order to reachfficiency10 while keeping inputs and environmental externalitiesonstant.

It should be noted that the DEA models of this paper presentariable returns to scale (VRS), thereby introducing a constrainthat makes the sum of the elements of the vector of intensitiesqual to one (Banker et al., 1984). There are various reasons forhe assumption of VRS. The first is that, the hypothesis that therexist VRS and a minimum efficient scale for each crop type isenerally accepted in agriculture. The second is that, since theeta-production frontier approach compares the efficiency of two

echnologies against the same efficiency frontier, the adopted effi-iency measure must be independent of the current productiontructures of the groups of firms being compared. In other words,he reference units being compared must be of the same size of thenit under analysis. Thus, absolute quantities of inputs and outputsatters as much as the output to environmental impact ratio in this

roblem. A constant returns to scale DEA model would set its effi-iency benchmark at the size of firm that has the capacity to achieve

relatively high output-to-environmental-pressure ratio, therebyreating a bias toward those groups of firms whose size distribu-ion is concentrated around that level. A VRS model, on the otherand, enables us to compare firms of the same size. Finally, whilehe calculation of economies of scale with undesirable outputs iselatively straightforward when using additive efficiency measuresSueyoshi and Goto, 2006), the radial nature of our measures makeshis estimation difficult to consider.

EA program for the estimation of individual Effg index

Overall technical efficiency, without environmental impact,ffg, is measured by means of a standard output-oriented radial

10 When working with Farrell output indices, the problem of slacks may arise.n efficiency score calculated with slacks in environmental restrictions means that

arms, having reached the maximum increase in output through an improvementn efficiency, are able to increase their environmental efficiency by reducing theirollutant emissions while keeping inputs and outputs constant. To check if slacks

n environmental restrictions distort efficiency indices, we use Assurance Regionnalysis (Thompson et al., 1990) to recalculate the Effsg scores. Results reveal a rela-

ionship between the aggregate Effsg index scores similar to that given by (2). Thus,rogram (2) holds. The efficiency scores given by the Assurance Region Analysis arevailable from the authors upon request.

se Policy 36 (2014) 275– 284 279

efficiency measure and VRS (Banker et al., 1984), using the follow-ing linear programming problem:

Effgk = min ˇ

subject to∑

k

�kynk ≥ ynk/ n = 1, . . ., N

∑

k

�kxmk ≤ xmk m = 1, . . ., M

�k ≥ 0 k = 1, . . ., K∑

k

�k = 1

(3)

In Appendix B we present the aggregated indices of these indi-vidual indices for the organic and conventional group of farms. And,in Appendix C we explain the bootstrap procedure. The estimationof the individual efficiency scores and the bootstrap analysis wereperformed using lpSolveAPI and FEAR software (Wilson, 2008) forplatform R.

Data and sample description

The sample used in this study comprises 8311 vineyard farmsfrom a single agro- climatic area of Navarre (Spain). 53 of thesefarms practice conventional agriculture and 30 are certified organic.The conventional farm data were taken from the 2004 FarmAccounting Data Network (FADN) and the organic farm data wereobtained from surveys of farmers, since the FADN has no availabledata on organic farms in this region. The organic farm sample rep-resents 60% of the total population of organic grape farms in theregion.

The variable set comprises three inputs, one output and two pol-lutants. The inputs are: land (hectares), labor (Annual Work Units),and other costs (fertilizers and pesticides, seeds, rent and deprecia-tion costs).12 Output is expressed as total farm revenues, exclusive ofsubsidies. We use monetary output values of instead of quantitiesto capture quality differences between organic and conventionalproducts. The quality differential of organic and conventional out-put is approximated by the price premium enjoyed by organicfarmers (Oude-Lansink et al., 2002). The average price premiumfor organic production in the sample is 16.7%.13

The farm-specific pollutant emission indicators were NitrogenSurplus (kg) and the Environmental Impact Quotient (EIQ) for thepesticides used by the farm. Nitrogen surplus values were esti-mated using the Soil Surface Balance Method (OECD, 2001). Thisstraightforward method takes the nitrogen cycle as the referenceand calculates the difference between the farm’s nitrogen inputsand outputs.

The pesticide toxicity index is based on the EIQ (Environmen-tal Impact Quotient) method developed by Kovach et al. (1992),which provides an average estimate of the toxic effect of each active

being dropped.12 Due to the sample size, aggregate data of capital depreciation and variable costs

were used. Pearson’s coefficient of correlation between these two variables in thesample is 0.61.

13 Fisher price index.

280 A.M. Aldanondo-Ochoa et al. / Land Use Policy 36 (2014) 275– 284

Table 1Descriptive statistics of the dataset.a

Land (Ha) Labor (AWU) Other cost (D ) Output (D ) Nitrogen (kg) EIQF (units)

Whole sample 55.22 (50.11) 1.61 (0.76) 18596.10 (12412.26) 53145.37 (35704.13) 5189.70 (5615.41) 1857.76 (809.83)Conventional 58.94 (52.29) 1.54 (0.55) 19537.31 (13064.93) 47084.40 (30763.08) 5219.04 (5088.73) 1545.12 (409.55)

1183.

oc

vlaabd

icp2dappbcpoot8mRcsebsh

R

tpwb

et

t(

mo

maE

i

increased by 14.3% (28.6%) each relative to its own efficiency fron-tier (without varying the quantity of inputs and the environmentalimpact of the farms). The organic farms are therefore closer to their

Table 2DEA aggregate efficiency scores, bias, standard deviation, and confidence intervals.a

DEA Est. Biasa S.D.a 95% CI boundsa

Lower Upper

OrganicEffg 0.784 0.065 0.052 0.628 0.828

Effse 0.857 0.057 0.041 0.734 0.894Prods 0.983 −0.006 0.021 0.966 1.042

Effsg 0.842 0.050 0.044 0.716 0.884Regulation 0.931 0.020 0.028 0.869 0.972ConventionalEffg 0.559 0.092 0.059 0.355 0.581

Effse 0.714 0.090 0.054 0.526 0.734Prods 0.899 0.010 0.044 0.824 0.992

Effsg 0.640 0.088 0.055 0.450 0.658Regulation 0.873 −0.010 0.061 0.782 1.009Conv./Org.Effg 0.713 0.056 0.077 0.499 0.803

Effse 0.833 0.048 0.072 0.645 0.922Prods 0.915 0.017 0.056 0.796 1.019

Effsg 0.760 0.059 0.073 0.559 0.841Regulation 0.938 −0.030 0.053 0.884 1.088

a 2000 bootstrap replications (Appendix C); (�c = 0.8, �o = 0.8).

Table 3DEA mean efficiency scores, bias, standard deviation, and confidence intervals.a

DEA Est. Biasa S.D.a 95% CI boundsa

Lower Upper

Organic 48.28 (45.6) 1.74 (1.04) 16933.29 (1

a Average values, standard deviations between parentheses.

f the active ingredients contained in the pesticide used14 by theorresponding farm.

Data on nitrogen and pesticide use were gathered via inter-iew with organic farmers. The pesticide and fertilizer usageevels of the conventional farms were estimated using the avail-ble FADN data and the cost of the representative fertilizationnd pesticide strategies for vineyards in Navarre recommendedy the Navarre Agricultural Technical Management Agency.15 Theescriptive statistics of the variables are given in Table 1.

While showing similar average values of land, labor and othernputs, the organic and conventional farms have what can beonsidered very low nitrogen surplus and EIQF levels when com-ared with other agricultural output sectors (Thomassen et al.,008, 2009). In the case of the nitrogen surplus values, this isue to the limited use of fertilizers16 for non-irrigated crops inrid areas. The emission rates are low in EIQF terms for bothroduction systems. Surprisingly, however, the organic farms dis-lay higher average values than the conventional farms, possiblyecause of their heavy reliance on sulphur and copper-based fungi-ides, which have higher toxicity levels than some conventionalesticides (Edwards-Jones and Howells, 2001). In line with EUrganic standards, sulphur and copper fungicides are permittednly in exceptional circumstances, such as an immediate threat tohe crop (Council Regulation (EC) 2092/91; Council Regulation (EC)99/2008). A restriction has also recently been placed on the per-itted amount of copper fungicide per year per hectare (Council

egulation (EC) 899/2008). The difficulty of waging biological pestontrol in an open farming system and the high costs involved inwitching to more plague-resistant varieties of grape may help toxplain why organic farmers rely mainly on copper and sulphur-ased pesticides. Nevertheless, none of the organic farms in theample reaches the permitted maximum limit of 6 kg of copper perectare.

esults

Tables 2 and 3 give, respectively, the DEA estimates ofhe technical efficiency, environmental efficiency, environmentalroductivity, and impact-of-regulation index scores (aggregatedeighted means in Table 2 and simple means in Table 3), plus the

ootstrap bias, standard deviations and confidence intervals.

Simar and Zelenyuk (2007) propose a Bootstrap test17 of differ-

nces in efficiency between groups of firms.18 The test statistic ishe quotient of the aggregate (or average) efficiency scores of the

14 More information available at: www.nysipm.cornell.edu/publications/eiq.15 We use the unitary cost of the fertilizer and the pesticide representative packageo derive physical data. This is a standard method of approximation in the literatureFäre et al., 2006; Morrison et al., 2002; Shaik et al., 2002).16 No farm was found to reach the 170 kg of nitrogen per ha limit on the use ofanure set by the Nitrates Directive (Council Directive (EEC), 676/91) and European

rganic farming regulations (Council Regulation (EC), 899/2008).17 It should be noted that the asymptotic distribution of the DEA efficiency esti-ator has been established for regular and compact technologies, with VRS (Simar

nd Wilson, 2008). The bootstrap method therefore has proven consistency for theffg, index and works as a sensitivity analysis for the rest of the indices.18 Simar and Wilson (1999) propose the use of a similar test for productivityndices.

98) 64431.31 (41703.63) 5135.08 (6582.22) 2439.91 (1031.31)

different group samples. If the Bootstrap confidence interval of thisquotient does not cover unity, the null hypothesis of equality inefficiency between groups is rejected. Tables 2 and 3 present thequotients of the group efficiency scores of the conventional andorganic farms.

The DEA aggregated technical environmental efficiency Effsescore is higher for organic farms (0.857) than for conventional farms(0.714), and this difference is significant at the 5% level. This meansthat the aggregate output of organic (conventional) farms can be

OrganicEffg 0.790 0.064 0.049 0.640 0.827

Effse 0.854 0.055 0.041 0.730 0.889Prods 0.986 −0.007 0.018 0.972 1.039

Effsg 0.844 0.048 0.043 0.721 0.889Regulation 0.938 0.016 0.025 0.882 0.978ConventionalEffg 0.546 0.075 0.055 0.362 0.576

Effse 0.732 0.075 0.048 0.567 0.756Prods 0.872 0.021 0.046 0.778 0.956

Effsg 0.625 0.055 0.051 0.434 0.630Regulation 0.869 −0.010 0.058 0.781 1.002Conv./Org.Effg 0.691 0.039 0.073 0.492 0.788

Effse 0.857 0.033 0.067 0.692 0.950Prods 0.884 0.028 0.056 0.754 0.974

Effsg 0.740 0.068 0.069 0.531 0.801Regulation 0.926 −0.026 0.054 0.863 1.069

a 2000 bootstrap replications (Appendix C); (�c = 0.8, �o = 0.8).

Land U

os

gttcf(ieTivsa

mmaai

toootoeu

etabrawfoliFdstm

eftasaa

evsa

oi

A.M. Aldanondo-Ochoa et al. /

wn efficiency frontier than the conventional farms are. Table 3hows similar results for the DEA mean efficiency scores.

Both DEA estimates of the environmental productivity (aggre-ate and simple means) are higher for the organic sample than forhe conventional farm sample. Aggregated (simple mean) Prods forhe organic sample is 0.983 (0.986), versus 0.899 (0.872) for theonventional sample. However, while the organic and conventionalarms do not differ significantly in their aggregate index scoreswhere the efficiency of each farm is weighted according to its sharen the aggregate output of its group), they show significant differ-nces in their mean scores (where all farms are equally weighted).his could be because the weighted Prods index of the organic groups practically the same as the mean score, whereas that of the con-entional group is slightly higher than the mean score. This resultuggests that the large farms in the conventional sample tend tochieve higher environmental productivity than the smaller ones.

Their higher technical environmental efficiency and environ-ental productivity scores show that the organic agriculture isore efficient in the use of natural resources. Thus, it is able to

chieve higher output than conventional agriculture for the samemount of inputs and pollutant emissions, as reflected by the Effsgndex in Tables 2 and 3.

Finally, it can be seen from Table 2 that the aggregated index ofhe impact of regulation (Regulation) score is close to unity on bothrganic (0.931) and conventional (0.873) farms. This means thatrganic farm sector sustain a reduction of the potential output lossf 6.9% and conventional farm sector an output loss of 12.7% in ordero reduce their pollutant emissions. This low impact of regulationn productivity may indicate that neither organic nor conventionalnvironmental standards prove restrictive in vineyards cultivatednder dry land conditions.

The results show, moreover, that there are no significant differ-nces in regulatory impact between the two systems. This suggestshat organic and conventional farms could achieve similar pollut-nt emission per unit input. There also appears to be no differencesetween large and small organic farming units. This lack of envi-onmental effectiveness in organic farming standards raises variousssumptions: (1) the extensive management practices associatedith conventional grape production are relatively environmentally

riendly and therefore not strikingly different from organic meth-ds, (2) the insufficient development of organic technology and theack of agronomic expertise of organic farmers might be undermin-ng the environment-enhancing potential of organic agriculture.or instance, since biological pest control technology for grape pro-uction is still under development in the region, farmers resort toulphur and copper fungicides, which are highly toxic. Likewise,his region’s farmers are ill-informed about the nutrient content of

anure and tend to apply it in excessive quantities.19

Finally, Tables 2 and 3 show that the DEA overall technicalfficiency score (Effg) for organic farming (0.784) is higher thanor conventional farming (0.559). This difference is significant athe 5% level and slightly higher than the price premium, which,ccording to our estimations, is around the 16.7% mark. This resultuggests that the organic grape farms in the sample produce highernd better-quality output than conventional farms using the samemount of inputs.

Overall, extensive vineyard cultivation is environmentally morefficient under an organic production system than under a con-

entional one, since it achieves higher levels of output with theame degree of nitrogen pollution and the same EIQ. This is largelyttributable to the higher technical environmental efficiency of

19 There has been a recent campaign to educate the region’s farmers in the usef simple methods to approximate the nutrient composition of manures and thusmprove their fertilization practices.

se Policy 36 (2014) 275– 284 281

organic farms, whose actual production, compared to that ofconventional farms, is closer to the potential for their specific tech-nology. The productivity differences are less clear. The aggregatedproductivity scores of the organic group are higher than those of theconventional group, but the difference is non-significant accordingto the tests used.

As far as environmental pressure is concerned, organic and con-ventional farming methods produce similar results. Both makeextensive use of natural resources, and conversion to organic pro-duction does not appear to deliver environmental gains: organicfarmers appear to make no greater output reduction than conven-tional farmers to control either nitrogen emissions or EIQ per unitof input. The evidence does nevertheless suggest a positive impacton indicators not considered in this paper, such as soil fertility,where improvements can be observed, since the overall technicalefficiency of organic farming is far superior to that of conventionalfarming. This may be due to the favorable impact of manure on soilstructure and water retention capacity.

In this respect, the overall technical efficiency index is an indi-cator of the additional advantages of organic farming methods forthe support of sustainable agriculture under arid conditions. Usingthe same amount of inputs, organic farmers are able to obtain moreproduce and command better prices than conventional farmers.

Discussion and limitations

Generally speaking, conversion to organic production meth-ods enables farmers to improve their environmental performance,especially per unit of land farmed, although there is evidence thatorganic agriculture has fewer advantages for extensive crop culti-vation in mountainous (De Boer, 2003) or arid areas (Guzmán andAlonso, 2008; Guzmán and González de Molina, 2009). Our findingsare in line with those of these recent studies. We have found that, inrelative terms, environmental gains from organic crop cultivationunder dry land conditions are low for both large and small farms,due to the extensive nature of both cultivation systems.

According to the “conventionalization” hypothesis, the indus-trialization and intensification of organic agriculture could have anegative impact on farms’ environmental performance. Although,under dry land conditions, the technical options for intensifica-tion are scant, results suggest that organic farmers focus moreon improving productive efficiency than on controlling pollutantemissions. They may be driven in this respect by the need to over-come the technological restrictions imposed by organic standardsand the uncertainty of using new technology.

Another of the issues addressed in this study is whether organicagriculture is more environmentally efficient than conventionalfarming. The results show that organic agriculture is more efficientbecause it produces higher output than conventional agriculturefor the same quantity of inputs and pollutant emissions. Thereare two reasons for this. The first is that organic farms are moreefficient than conventional farms, because, using their own tech-nology set, they come closer to their own production frontier. Thisfinding could be due, in part, to the fact that our sample containsfewer organic than conventional farms. Nevertheless, the higherefficiency of organic farms with respect to their own technologyfrontier is a finding that has emerged repeatedly in studies involv-ing other crops (Oude-Lansink et al., 2002; Tzouvelekas et al., 2001).In this study, we find organic farms to be more efficient even whenthe analysis is inclusive of environmental externalities. The sec-ond reason is that organic technology is no less productive than

conventional technology.

Thus, organic farmers achieve higher output from the samequantity of inputs. This finding is consistent with previous studiesthat have shown organic farming to produce higher yields than

2 Land U

c(

rclwotv(

teb

aTccnisstdtr

C

nndeaittdmr

svpvtttuobfun

ftI(tu

tional technical efficiency frontier, net of environmental impact.Ri-Rii-R4-Riii is the corresponding organic technical efficiency fron-tier. Ci-Cii-C3-Riii is the metafrontier. Fig. A1 depicts the inefficientorganic unit, C, shown in Figs. 1 and 2. This unit uses the x inputs

82 A.M. Aldanondo-Ochoa et al. /

onventional farming in agriculturally less-endowed areasBadgley et al., 2007; Lotter, 2003).

One of the limitations of this study is that it considers two envi-onmental indicators, nitrogen surplus and pesticide toxicity, whileapturing soil fertility only indirectly. The unavailability of farm-evel data has prevented us from considering other key indicators in

hich organic farming performs well. Nevertheless, the findings ofur study are robust for the pollutants considered and coincide withhe results of other studies comparing extensive organic and con-entional crop cultivation systems based on alternative indicatorsSchader, 2009; Zimmermann et al., 2011).

Furthermore, despite the fact that farm productivity is sensitiveo technological change, this study uses input–output and pollutantmission observations for one year only. This limitation remains toe addressed in future research.

Finally, the DEA method is particularly sensitive to dimension-lity problems (small sample with multiple inputs and outputs).his can lead to inaccurate estimates with high variance and highonfidence levels. In this study, as in most of the research on the effi-iency of organic farming practices, we have worked with a limitedumber of observations because the target population for this crop

n this agro-climatic area is small, as stated in the methodologyection. Nevertheless, the confidence intervals of the estimates areufficient to enable comparison of the efficiency scores. Moreover,he results still hold when the bootstrap analysis is repeated usingifferent-sized subsamples (� i = 0.7, � i = 0.9, i = c.o). In addition, fur-her tests20 on differences in group efficiency scores show similaresults.

onclusions and policy implications

The empirical findings of this study apply to particular agro-omic conditions: extensive production of vine in semi-arid,on-irrigated areas. The results suggest that, under these con-itions, organic production systems are environmentally morefficient than conventional systems because they are able tochieve higher output for the same degree of environmentalmpact and quantity of inputs. In addition, organic agricul-ure produces higher output than conventional agriculture forhe same quantity of inputs under the above-mentioned con-itions. Organic cultivation could therefore provide an efficienteans to maintain environmentally-sustainable agriculture in arid

egions.As far as the environmental effectiveness of organic farming

tandards is concerned, the results of this study suggest that con-entional farms perform on a par with organic farms in terms ofollutant emission control per unit input in extensive crop culti-ation of vine in arid areas. We find no significant difference inhe impact of environmental regulation on productivity betweenhe two systems, which might indicate that organic and conven-ional farms stand more or less equal in terms of emissions pernit input. Thus, in relative terms, the environmental gains fromrganic vineyard cultivation under dry land conditions appear to

e low. This impression may stem from the fact that this studyocuses on extensive vine cultivation, where emissions are so lownder both production systems that restrictions on the use ofitrogen or copper per hectare are not enforced. It is therefore

20 The efficiency scores were computed using the FDH (Free Disposal Hull) methodor the maximum number of variables and the smallest sample and over 50% ofhe observations were classed as inefficient. This mitigated the sample-size bias.n addition, the application of the two-step method developed by Daraio and Simar2007) to the Effg index shows that the differences between the two systems retainedhe same level of significance. The results of these tests are available from the authorspon request.

se Policy 36 (2014) 275– 284

worth investigating this issue for other crops and agro-climaticareas.

Overall, our results suggest that organic vine dry land farm-ing achieves a higher output than conventional dry land farmingwith a similar environmental impact. One of the advantages of theresearch method proposed in this study is that it enables the sepa-ration of purely technical issues from managerial performance andefficiency issues in the analysis of environmental effectiveness andefficiency. This separation is important when considering possibleregulatory reform or a change of policy with respect to grants fororganic agriculture.

One of the proposals for improving the environmental per-formance of organic farms is to tighten up environmentalregulations. This proposal is related to the “conventionalization”hypothesis. However, under dry land conditions, the techni-cal options for intensification and industrialization of organicagriculture are scant. And, in this paper we show that theenvironmentally-kinder nature of extensive vineyard cultiva-tion under either production system, be it organic or con-ventional, may provide an argument against stricter organicstandards for farmers working on the production of vine in aridland.

Grants for organic farmers engaged in extensive vine dry-landcrop cultivation would improve the environmental efficiency ofagriculture. This, together with the fact that organic farmers attainmore output for the same amount of inputs, suggest that organiccultivation could be an efficient system for maintaining sustain-able agriculture in arid regions. However, we have found thatorganic vine farmers have achieved better results increasing out-put efficiency per environmental impact than controlling pollutantemissions per unit of input. This might suggest that, in this specificcase of extensive vine cultivation could be convenient to addresspart of organic subsidies to further improvements in the control ofpollution from fertilizers and pesticides.

Appendix A.

Fig. A1 shows the production curves of two technologies thatuse x inputs to produce y outputs. It illustrates the overall technicalefficiency defined in Fig. 2. In Fig. A1, Ci-Cii-C3-Civ is the conven-

Fig. A1. Input–output production frontiers with two technologies and without envi-ronmental impacts.

Land U

nFatxbN

A

educerctmopessuf

csiro

wetˇtyat

haerAis

A

c(yim(

A.M. Aldanondo-Ochoa et al. /

ot shown in Figs. 1 and 2. All the production units represented inigs. 1 and 2 lie on the line x-C3 in Fig. A1. We also depict units R4nd C3 from Figs. 1 and 2, that is, the organic and conventional unitshat achieve the highest output Y for the same amount of inputs,. The overall conventional technical efficiency of unit C is giveny the ratio between segments xC/xC3, which is equal to the ratioC/NG in Fig. 2.

ppendix B. Aggregation of individual indices

One of the advantages of aggregate efficiency indices is that theirstimates illustrate the efficiency of an entire production set or ofifferent groups of firms within it. There were several reasons forsing aggregate indices in this paper to compare organic versusonventional agricultural performance. Firstly, a unique efficiencystimate for each system could provide key information for envi-onmental policy designers. Secondly, it enables us to make ouromparison of the performance of these two agricultural produc-ion systems using the Simar and Zelenyuk (2007) method. This

ethod is, as we are about to see, has the advantage of being basedn bootstrap resampling, which is a technique to extract subsam-les from the original sample and repeat all the efficiency indicatorstimations for each of these subsamples. Therefore, with balancedubsampling, we test the sensitivity of the aggregated efficiencycores to variations in the firm sample composition. This is partic-larly useful for analysis in sectors with small samples, a problemrequently encountered in organic farming research.

This study uses the inverse of the industry output-oriented effi-iency index (Färe and Zelenyuk, 2003, 2007; Farrell, 1957) foringle-product firms in order to aggregate all the indices. The result-ng aggregate index of the Effg, Effse and Effsg efficiency indicesepresents the ratio of aggregate efficient output to aggregate actualutput, and is given by the following expression:

EBi =∑k=Ki

k=1 yik∑k=Kik=1 ye

ik

=k=Ki∑

k=1

ˇikSeik

Seik

= yeik∑k=Ki

k=1 yeik

(4)

here EBi denotes indistinctly the Effse, Effsg or Effg, aggregatedfficiency index of the farming ith subsector (organic or conven-ional), yik is the output of the kth farm in the ith farming subsector;ik is each farm’s individual corresponding efficiency score; ye

ikis

he efficient output of the kth farm the ith farming subsector, theneik

= yik/ˇik. Then, Seik

denotes the share of the kth farm in the over-ll technically-efficient output of each farming subsector; and Ki ishe number of farms in each subsector.

The aggregated Prods index and the aggregated Regulation indexave been calculated using a similar procedure. In the case of theggregate Prods the index is weighted by the overall environmentalfficient output ye

ik= yik/Effsgik, where Effsgik is the overall envi-

onmental technical efficient score of the kth firm in the ith sector.nd, in the aggregated Regulation index ye

ik= yik/Effgik, where Effgik

s the overall technical efficiency score of the kth firm in the ithector.

ppendix C. The bootstrap procedure

Bootstrapping is a data simulation method in which statisti-al inference is used to determine the accuracy of the estimatesEfron and Tibshirani, 1993). It has been used in sensitivity anal-

sis and in statistical inference (Simar and Wilson, 2000), and tomprove the robustness of non-parametric efficiency estimation

ethods (Simar and Wilson, 2008) such as DEA. Simar and Zelenyuk2007) were the first to use the bootstrap technique to estimate the

se Policy 36 (2014) 275– 284 283

sensitivity of aggregate DEA efficiency scores to variations in sam-ple composition. They used balanced subsampling (Kneip et al.,2008) to estimate the bias and confidence intervals of aggregateefficiency scores and to test for differences in efficiency betweendifferent groups of firms. The procedure adopted in this study isthe same as in Simar and Zelenyuk (2007) and can be describedsimply as follows.

2000 bootstrap replications are made by drawing one subsam-ple from the conventional farm sample and another from the initialorganic farm sample. Estimates of the previously defined aggregateefficiency scores are calculated for every replication. The bootstrapefficiency scores are used to compute the bias and the confidenceinterval of the aggregate DEA and to test for efficiency differencesbetween groups. The appropriate size for the subsample used toobtain bootstrap estimates is determined by the following rule(Kneip et al., 2008; Simar and Wilson, 2009):

ri = R−�ii ,

and

�i = R−1/3(n+m+1)i

where ri = the size of the subsample from group i; Ri = group I’s ini-tial sample size; and n and m are the numbers of outputs and inputs,respectively, used by each unit in the sample.

References

Arandia, A., Aldanondo-Ochoa, A., 2008. Social versus Private Efficiency: A Compar-ison of Conventional and Organic Farming Systems in Vineyard Production. In:European Association of Agricultural Economists International Congress, August26–29, Ghent, Belgium.

Arandia, A., Aldanondo-Ochoa, A., 2011. Pollution shadow prices in conventional andorganic farming: an application in a Mediterranean context. Spanish Journal ofAgricultural Research 9 (2), 363–376.

Backer, D.E., Aertsens, J., Vergucht, S., Steurbaut, W., 2009. Assessing the ecologi-cal soundness of organic and conventional agriculture by means of life cycleassessment (LCA). A case study of leek production. British Food Journal 111 (10),1028–1061.

Badgley, C., Moghtader, J., Quintero, E., Zakem, E., Chappell, M.J., Avilés-Vázquez, K.,Samulon, A., Perfecto, I., 2007. Organic agriculture and the global food supply.Renewable Agriculture and Food Systems 22 (2), 86–108.

Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimating technicaland scale inefficiencies in Data Envelopment Analysis. Management Science 30,1078–1092.

Battese, G.E., Prasada Rao, D.S., Donnell, O.C.J., 2004. A metafrontier productionfunction for estimation of technical efficiencies and technology gaps for firmsoperating under different technologies. Journal of Productivity Analysis 21,91–103.

Buck, D., Getz, C., Guthman, J., 1991. From farm to table: the organic veg-etable commodity chain of Northern California. Sociologica Ruralis 37 (1),3–20.

Casey, J.W., Holden, N.M., 2006. Greenhouse gas emissions from conventional,agri-environmental scheme, and organic Irish suckler-beef units. Journal of Envi-ronmental Quality 35 (1), 231–239.

Cederberg, C., Mattsson, B., 2000. Life cycle assessment of milk production: a com-parison of conventional and organic farming. Journal of Cleaner Production 8(1), 49–60.

Chen, W.C., Johnson, A.L., 2010. A unified model for detecting efficient and inefficientoutliers in data envelopment analysis. Computers and Operation Research 37,417–425.

Council Directive (EEC) No 676/91 of 12 December 1991 concerning the pro-tection of waters against pollution caused by nitrates from agriculturalsources.

Council Regulation (EEC) No 2092/91 of 24 June 1991 on organic production of agri-cultural products and indications referring thereto on agricultural products andfoodstuffs. OJ L 198, 22/07/1991.

Council Regulation (EC) No 834/2007 of 28 June 2007 on organic production andlabelling of organic products and repealing Regulation (EEC) No 2092/91. OJ L189, 20/07/2007.

Council Regulation (EC) No 889/2008 of 5 September 2008 laying down detailedrules for the implementation of Council Regulation (CE) no 834/2007 on organic

production and labelling of organic products with regard to organic production,labelling and control. OJ L 250, 18/9/2008.

Daraio, C., Simar, L., 2007. Conditional nonparametric frontier models for convex andnon-convex technologies: a unifying approach. Journal of Productivity Analysis28, 13–32.

2 Land U

D

D

D

E

E

F

F

F

F

F

F

F

G

G

G

H

H

H

H

K

K

K

K

K

K

L

M

M

M

O

O

O

system in China: a data envelopment analysis approach. Ecological Economics68, 306–316.

84 A.M. Aldanondo-Ochoa et al. /

arnhofer, I., Lindenthal, T., Bartel-Kratochvil, R., Zollitsch, W., 2010. Conven-tionalisation of organic farming practices: from structural criteria towards anassessment based on organic principles: a review. Agronomy for SustainableDevelopment 30, 67–81.

e Boer, I.J.M., 2003. Environmental impact assessment of conventional and organicmilk Production. Livestock Production Science 80 (1–2), 69–77.

yckhoff, H., Allen, K., 2001. Measuring ecological efficiency with data envel-opment analysis (DEA). European Journal of Operational Research 132,312–325.

dwards-Jones, G., Howells, O., 2001. The origin and hazard of inputs to crop pro-tection in organic farming systems: are they sustainable? Agricultural Systems67 (1), 31–47.

fron, B., Tibshirani, R.J., 1993. An Introduction to the Bootstrap. Chapman and Hall,London.

äre, R., Grosskopf, S., 2003. Nonparametric productivity analysis with undesir-able outputs: comment. American Journal of Agricultural Economics 85 (4),1070–1074.

äre, R., Grosskopf, S., Lovell, C.A.K., Pasurka, C., 1989. Multilateral productivity com-parisons when some outputs are undesirable: a nonparametric approach. TheReview of Economics and Statistics 71, 90–98.

äre, R., Grosskopf, S., Pasurka, C., 2007. Pollution abatement activities and tradi-tional productivity. Ecological Economics 62 (3–4), 673–682.

äre, R., Grosskopf, S., Weber, W.L., 2006. Shadow prices and pollution costs in USagriculture. Ecological Economics 56, 89–103.

äre, R., Zelenyuk, V., 2003. On aggregate Farrell efficiencies. European Journal ofOperational Research 146, 615–620.

äre, R., Zelenyuk, V., 2007. Extending Färe and Zelenyuk (2003). European Journalof Operational Research 179, 594–595.

arrell, M.J., 1957. The Measurement of Productive Efficiency. Journal of the RoyalStatistic Society 120, 449–460.

uthman, J., 2004. The trouble with ‘Organic Lite’ in California: a rejoinder to the‘conventionalisation’ debate. Sociologia Ruralis 44 (3), 301–316.

uzmán, G.I., Alonso, A.M., 2008. A comparison of energy use in conventionaland organic olive oil production in Spain. Agricultural Systems 98 (3),167–176.

uzmán, G.I., González de Molina, M., 2009. Preindustrial agriculture versusorganic agriculture. The land cost of sustainability. Land Use Policy 26 (2),502–510.

aas, G., Wetterich, F., Kopke, U., 2001. Comparing intensive, extensified and organicgrassland farming in southern Germany by process life cycle assessment. Agri-culture, Ecosystems and Environment 85, 43–53.

ailu, A., Veeman, T.S., 2001. Non-parametric productivity analysis with undesir-able outputs: an application to the Canadian pulp and paper industry. AmericanJournal of Agricultural Economics 83 (3), 605–616.

ayami, M., Ruttan, V.W., 1971. Agriculture Development in an International Per-spective. John Hopkins Press, Baltimore.

ole, D.G., Perkins, A.J., Wilson, J.D., Alexander, I.H., Grice, P.V., Evans, A.D., 2005.Does organic farming benefit biodiversity? Biological Conservation 122 (1),113–130.

neip, A., Simar, L., Wilson, P.W., 2008. Asymptotics and consistent bootstrapsfor DEA estimators in nonparametric frontier models. Econometric Theory 24,1663–1697.

orsaeth, A., Eltun, R., 2000. Nitrogen mass balances in conventional, integratedand ecological cropping systems and the relationship between balance calcula-tions and nitrogen runoff in an 8-year field experiment in Norway. Agriculture,Ecosystems & Environment 79 (2–3), 199–214.

ortelainen, M., Kuosmanen, T., 2007. Eco-efficiency analysis of consumer durablesusing absolute shadow prices. Journal of Productivity Analysis 28, 57–69.

ovach, J., Petzoldt, C., Degni, J., Tette, J., 1992. A method to measure the environ-mental impact of pesticides. In: New York’s Food and Life Sciences Bulletin 139.Cornell University, Ithaca, New York.

umbhakar, S., Tsionas, E., Timo Sipiläinen, E.G.T., 2009. Joint estimation of technol-ogy choice and technical efficiency: an application to organic and conventionaldairy farming. Journal of Productivity Analysis 31, 151–161.

uosmanen, T., Kortelainen, M., 2005. Measuring eco-efficiency of production withData Envelopment Analysis. Journal of Industrial Ecology 9, 59–72.

otter, D.W., 2003. Organic agriculture. Journal of Sustainable Agriculture 21 (4),59–128.

ansfield, B., 2004. Organic views of nature: the debate over organic certificationfor aquatic animals. Sociologia Ruralis 44 (2), 216–232.

ayen, C.D., Balagtas, J.V., Alexander, C.E., 2010. Technology adoption and technicalefficiency: organic and conventional dairy farms in the United States. AmericanJournal of Agricultural Economics 92 (1), 181–195.

orrison, J.C., Ball, V.E., Felthoven, A.G., Nhering, R.E., 2002. Effective costsand chemical use in United States agricultural production: using the envi-ronment as a free input. American Journal of Agricultural Economics 84,902–915.

’Donnell, C.J., Prasada Rao, D.S., Battese, G.E., 2008. Metafrontier frameworks forthe study of firm-level efficiencies and technology ratios. Empirical Economics34, 231–255.

ECD, 2001. OECD National Soil Surface Nitrogen Balances. Explanatory Notes. OECD

General Secretariat, Paris.

ECD, 2010. Guidelines for Cost-effective Agri-environmental Policy Measures.OECD General Secretariat, Paris.

se Policy 36 (2014) 275– 284

Offermann, F., Nieberg, H., Katrin Zander, H.K., 2009. Dependency of organic farmson direct payments in selected EU member states: today and tomorrow. FoodPolicy 34 (3), 273–279.

Oude-Lansink, A.K., Pietola, K., Bäckman, S., 2002. Efficiency and productivity ofconventional and organic farms in Finland 1994–1997. European Review ofAgricultural Economics 29, 51–65.

Padel, S., Röcklinsberg, H., Schmid, O., 2009. The implementation of organic prin-ciples and values in the European Regulation for organic food. Food Policy 34,245–251.

Picazo-Tadeo, A.J., Prior, D., 2009. Environmental externalities and efficiency mea-surement. Journal of Environmental Management 90 (11), 3332–3339.

Pimentel, D., Hepperley, P., Hanson, J., Douds, D., Seidel, R., 2005. Environmental,energetic, and economic comparisons of organic and conventional farming sys-tems. BioScience 55 (7), 573–582.

Pittman, R.W., 1983. Multilateral productivity comparisons with undesirable out-puts. Economic Journal 93, 883–891.

Rahmann, G., 2011. Biodiversity of organic farming: what do we know? Agricultureand Forestry Research 61 (3), 189–208.

Reed, M., 2005. The socio-geographies of organic farming and “convention-alisation”: an examination through the academic road trip. In: Paperpresented at the XXI Congress of the ESRS, vols. 22–27, August, Keszthely,Hungary.

Reinhard, S., Lovell, C.A.K., Thijssen, G., 1999. Econometric estimation of technicaland environmental efficiency: an application to Dutch dairy farms. AmericanJournal of Agricultural Economics 81 (1), 44–60.

Sauer, J., Park, T., 2009. Organic farming in Scandinavia: productivity and marketexit. Ecological Economics 68 (8), 2243–2254.

Schader, C., 2009. Cost-effectiveness of organic farming for achieving environmentalpolicy targets in Switzerland. Ph.D. thesis. Institute of Biological, Environ-mental and Rural Sciences, Aberystwyth University, Wales, Research Instituteof Organic Agriculture FiBL, Available at: https://www.fibl-shop.org/shop/(accessed 16.01.12).

Shaik, S., Helmers, G.A., Langemeier, M.R., 2002. Direct and indirect shadow priceand cost estimates of nitrogen pollution abatement. Journal of Agricultural andResource Economics 27, 420–432.

Simar, L., 2003. Detecting outliers in frontier models: a simple approach. Journal ofProductivity Analysis 20 (3), 391–424.

Simar, L., Wilson, P.W., 1999. Estimating and bootstrapping Malmquist indices. Euro-pean Journal of Operational Research 115, 459–471.

Simar, L., Wilson, P.W., 2000. Statistical inference in nonparametric fron-tier models: the state of the art. Journal of Productivity Analysis 13,49–78.

Simar, L., Wilson, P.W., 2008. Statistical inference in nonparametric frontier models:recent developments and perspectives. In: Fried, H., Lovell, C.A.K., Schmidt, S.S.(Eds.), The Measurement of Productive Efficiency. , 2nd edition. Oxford Univer-sity Press, Oxford, pp. 421–521.

Simar, L., Wilson, P.W., 2009. Inference by subsampling in nonparametric frontiermodels. Université Catholique de Louvaine. Institut de Statistique, DiscussionPaper, 0933.

Simar, L., Zelenyuk, V., 2007. Statistical inference for aggregates of Farrell-type effi-ciencies. Journal of Applied Econometrics 22, 1367–1394.

Stolze, M., Piorr, A., Haring, A., Dabbert, S., 2000. The environmental impacts oforganic farming in Europe. Organic Farming in Europe: Economics and Policy,vol. 6. University of Hohenheim, Stuttgart.

Sueyoshi, T., Goto, M., 2006. Measurement of Returns to Scale and Damages toScale for DEA-based operational and environmental assessment: how to managedesirable (good) and undesirable (bad) outputs? European Journal of Opera-tional Research 211 (1), 76–89.

Thomassen, M.A., van Calker, K.J., Smits, M.C.J., Iepema, G.L., de Boer, I.J.M., 2008.Life cycle assessment of conventional and organic milk production in theNetherlands. Agricultural Systems 96 (1–3), 95–107.

Thomassen, M.A., Dolman, M.A., van Calker, K.J., de Boer, J.M., 2009. Relating life cycleassessment indicators to gross value added for Dutch dairy farms. EcologicalEconomics 68 (8–9), 2278–2284.

Thompson, R.G., Langemeier, L.N., Lee, C.-T., Lee, E., Thrall, R.M., 1990. The role of mul-tiplier bounds in efficiency analysis with application to Kansas farming. Journalof Econometrics 46, 93–108.

Tuomisto, H.L., Hodge, I.D., Riordan, P., Macdonald, D.W., 2012. Does organic farmingreduce environmental impacts? – A meta-analysis of European research. Journalof Environmental Management 112 (15), 309–320.

Tzouvelekas, V., Pantzios, C.J., Fotopoulos, C., 2001. Technical efficiency of alterna-tive farming systems: the case of Greek organic and conventional olive-growingfarms. Food Policy 26, 549–569.

Wilson, P.W., 1993. Detecting outliers in deterministic nonparametric frontier mod-els with multiple outputs. Journal of Business and Economic Statistics 11 (3),319–323.

Wilson, P.W., 2008. FEAR 1.0: a software package for frontier efficiency analysis withR. Socio-Economic Planning Sciences 42, 247–254.

Zhang, B., Fan, J., Bi, Z.W., Yuan, Z.Y., Ge, J.J., 2008. Eco-efficiency analysis of industrial

Zimmermann, A., Baumgartner, D., Nemecek, T., Gaillard, G., 2011. Are public pay-ments for organic farming cost-effective? Combining a decision-support modelwith LCA. The International Journal of Life Cycle Assessment 16 (6), 548–560.