Land Use Policy 79 (2018) 375–385

Contents lists available at ScienceDirect

Land Use Policy

journal homepage: www.elsevier.com/locate/landusepol

Let the cows graze: An empirical investigation on the trade-off between T

efficiency and farm animal welfare in milk production
⁎
Hinrich D. Schultea, , Linda Armbrechtb, Rasmus Bürgerc, Matthias Gaulyd, Oliver Musshoffa,
Silke Hüttele
a
Farm Management Group, Department of Agricultural Economics and Rural Development, Faculty of Agricultural Sciences, University of Goettingen, Platz der Göttinger
Sieben 5, 37073 Göttingen, Germany
b
Livestock Production System Group, Department of Animal Science, Faculty of Agricultural Sciences, University of Goettingen, Albrecht-Thaer-Weg 3, 37075 Göttingen,
Germany
c
Thaer-Institute, Faculty of Life Sciences, Humboldt Universität zu Berlin, Hannoversche Straße 27, 10115 Berlin, Germany
d
Faculty of Science and Technology, Free-University of Bolzano, Piazza Università 5, 39100 Bolzano, Italy
e
Production Economics Group, Institute of Food and Resource Economics, Bonn University, Meckenheimer Allee 174, 53115 Bonn, Germany

A R T I C LE I N FO A B S T R A C T

JEL classification: To investigate whether farm animal welfare comes at the cost of dairy farm performance, and the role that
D24 pasture-access thereby plays, we analyse a rich sample of 45 dairy farms in Germany with a scientific measure of
Q12 farm animal welfare. Based on directional efficiency measure that acknowledges sequential preferences such that
Q18 farm animal welfare becomes relevant after technical efficiency, we cannot find a trade-off between farm animal
Keywords: welfare and technical milk efficiency. Pasture-based production systems can be at least as efficient or even more
Farm animal welfare efficient compared to confinement systems, despite lower milk yields. Neglecting sequential preferences would
Technical efficiency bias efficiency in provision of farm animal welfare by pasture-systems. Farms from all types of dairy systems
Pasture
determine the efficient frontier, where efficiency increases with herd size, and is linked with higher profits. We
Milk production
conclude that pasture-access may help but does not guarantee higher levels of animal welfare. These results
Data envelopment analysis
question governmental support for pasture access regarding the provision of farm animal welfare.

1. Introduction argumentation, in complex production systems such as dairy, the most

Farm animal welfare (FAW) has gained considerable attention in
recent decades, with a notable influence on consumers’ food choices
(Lagerkvist and Hess, 2011). The majority of studies reveal that con-
sumers’ willingness to pay (WTP) increases with higher standards of
farm animal welfare (e.g., Kehlbacher et al., 2012; Napolitano et al.,
2008). Interestingly, this also holds true for dairy products produced in
pasture-based systems. Consumers perceive pasture access as a positive
contributor to welfare (Cardoso et al., 2016; Ellis et al., 2009) and even
use it as a criterion to assess the level of FAW of the production system
(Schuppli et al., 2014). Access to pasture is in fact used as an important
sales argument (Weinrich et al., 2014), although the specific role of
pasture access on dairy cow welfare is not that straightforward (Arnott
et al., 2017). Contrarily, Lusk and Norwood (2011) argue that
FAW directly relates to the husbandry system. Based upon their
profitable milk yield, as a measure of productivity, does not accompany
optimal levels of FAW. Against the backdrop that many consumers find
existing husbandry practices inacceptable (Roosen et al., 2016), several
policy measures to incentivise production at higher levels of FAW, in-
cluding pasture access for cows, have emerged (Ingenbleek et al., 2012;
Veissier et al., 2008). The idea is that farms receive compensation since
pasture access requires sacrificing (milk) productivity and thus profit-
ability to provide higher levels of FAW. The question whether simply
providing pasture-access warrant payments because farms have to sa-
crifice milk yield and profitability to increase dairy cows’ welfare,
however, remains open.
From animal science perspective, most animals prefer to stay on
pasture, if cows are given the choice between pasture or confinement in
an experimental setting (Charlton et al., 2011; Legrand et al., 2009;
Motupalli et al., 2014). Studies in the field of animal science, however,

⁎
Corresponding author.
E-mail addresses: hinrich.schulte@agr.uni-goettingen.de (H.D. Schulte), linda.armbrecht@agr.uni-goettingen.de (L. Armbrecht),
rasmus.buerger@agrar.hu-berlin.de (R. Bürger), matthias.gauly@unibz.it (M. Gauly), oliver.musshoff@agr.uni-goettingen.de (O. Musshoff),
s.huettel@ilr.uni-bonn.de (S. Hüttel).

https://doi.org/10.1016/j.landusepol.2018.07.005
Received 30 January 2018; Received in revised form 2 July 2018; Accepted 3 July 2018
0264-8377/ © 2018 Published by Elsevier Ltd.
H.D. Schulte et al.
report mixed findings concerning specific health and welfare effects of
pasture access. Benefits result from more space, and easily accessible
lying opportunities, which increases the lying times of cows, leading to
lower levels of aggression (e.g., Burow et al., 2011). Furthermore, re-
duced integument alterations, hair loss, lesions and swellings have been
reported for cows with pasture access (Burow et al., 2013). While some
authors report positive health effects related to reduced lameness
(Barker et al., 2010; Chapinal et al., 2013; de Vries et al., 2015), others
could not determine any effects on claw health (Baird et al., 2009;
Chapinal et al., 2010). Moreover, high-yielding cows might suffer from
insufficient energy intake with pasture feeding, which in turn leads to
nutritional and metabolic stress, with a negative impact on welfare
(Vance et al., 2012). Beyond this, parasites that accompany grazing
might also reduce positive impacts on FAW (Bennema et al., 2011;
Vanderstichel et al., 2012).
From production economics perspective, dairy farm performance
has often been examined using technical, economic and environmental
efficiency (cf. among others Breustedt et al., 2011; Wettemann and
Latacz-Lohmann, 2017). Only few studies directly relate farm (tech-
nical) efficiency to cow health status, as one aspect of welfare, and
report a positive relationship (e.g., Barnes et al., 2011; Lawson et al.,
2004a). The specific role of pasture access in this regard has thus far
received only little attention. The study of Allendorf and Wettemann
(2015) forms one exception and does not support any relationship be-
tween dairy cow welfare, pasture access and performance. These stu-
dies, however, rely on lameness or somatic cell counts as simplistic
proxies for dairy cows’ welfare and thus neglect the high complexity
and multi-dimensionality of animal welfare. Furthermore, these studies
do not consider FAW as a central part of the production process, as is
argued by McInerney (2004), and simply relate efficiency scores to
health status or pasture access indicators. In a recent study Henningsen
et al. (2018) acknowledge that animal welfare can in fact positively
impact the production process and its performance, though their em-
pirical analysis reveals only a weak relationship between welfare and
performance for Danish pig farmers. Evidence for the dairy sector that
acknowledges the complexity of farm animal welfare and peculiarities
under pasture-access is thus far not available.
Our study aims to close this gap by empirically investigating the
relationship between dairy farm performance, pasture access of dairy
cows and their welfare status. The objectives are twofold. Firstly, we
empirically investigate the potential trade-off between dairy farm per-
formance and FAW. Secondly, we analyse whether this potential trade-
off differs systematically due to pasture access. We refer to technical
efficiency as one major dimension of performance besides economic
efficiency, productivity and effectiveness.1 Our study relies on a rich
dataset of 45 dairy farms located in north-western Germany, where in
addition to all dairy branch-specific costs, a sophisticated measure to
assess dairy cow welfare based on the Welfare Quality® Protocol (WQP)
for cattle (Welfare Quality®, 2012) is available. Our study directly re-
lates to research by Gocsik et al. (2016), who used the WQP-index in
their analysis on the cost-efficiency of improvements in broiler welfare.
In contrast and following Asmild and Hougaard (2006), we refer to a
directional efficiency measure that acknowledges behavioural patterns
of farm managers with sequential preferences: improvements in pure
technical efficiency might be the primary goal of the farm and after
having achieved the desired level of efficiency, farm animal welfare
might become the next relevant focus. Therefore, in this study, we
consider farm animal welfare as a second output alongside milk and are
1 (2011) found healthier herds with fewer incidences of lameness were
Technical efficiency is one major dimension of performance besides eco-
nomic efficiency, productivity and effectiveness. While productivity usually
describes the relationships between all inputs and outputs at the farm level,
efficiency is a normative measure of specific input-output relations.
Effectiveness is also a normative measure used to capture whether an intended
output level is produced, irrespective of the current efficiency level.
376
Land Use Policy 79 (2018) 375–385
the first to use the multi-disciplinary scientific animal welfare measure
within an efficiency analysis framework. To cross-validate the as-
sumption of sequential preferences we further provide results of a
standard efficiency approach. Our study will help to inform the dis-
cussion on whether policy support for pasture access can be justified by
improved animal welfare, and whether pasture-based systems can ef-
ficiently provide farm animal welfare.
The remainder of this paper is structured as follows: first, we review
the relevant literature and present the analytical framework (Section 2).
Then, we describe the dataset and how farm animal welfare is measured
based on the WQP (Section 3). In Section 4, we present the empirical
strategy and in Section 5, we present and discuss the results. In the final
section, we conclude.
2. Background, related literature and research framework
The percentage of dairy farms in Europe with pasture access
dropped from 52% in 2008 to 35% in 2012 (Reijs et al., 2013). This
trend contradicts consumers’ increased interest in pasture-based sys-
tems but is the result of higher economic pressure induced by less stable
and lower output prices, as well as the EU’s milk market policy, which
features reduced market support. Cost pressure might lead to changes in
feeding strategies with tendencies to more maize-based rations given
the higher energy yield per hectare compared to grassland. In addition,
benefits of higher technical efficiency in milk production on specialized
farms have been reported (e.g., Pieralli et al., 2017). Farms with larger
herds might also benefit from economies of scale. Trends for larger
herds and more specialization might partly explain the trend in pasture
use in Western Europe since providing pasture access for larger herds
can be more demanding on the grassland itself (given a confined
grazing area). Against this backdrop, several policy schemes of the
second pillar of the EU’s Common Agricultural Policy have been de-
voted to counteract the trends in grassland and pasture-use. These
measures target at keeping grassland in production to maintain land-
scapes, functioning of ecosystems, but also enhancing biological di-
versity and farm animal welfare. The latter programmes’ success is
often hampered by low acceptance rates among farmers. Schreiner and
Hess (2017) argue that the high personal animal welfare standards of
farmers are often mismatched with the design of such programmes.
Farmers might fear disutility from participation, for instance, by suf-
fering a loss of autonomy. This, however, does not necessarily imply
that farmers do not include farm animal welfare in their goal setting.
Hansson and Lagerkvist (2015) distinguish in this context between use
and non-use. Use values are directly related to productivity and denote
the necessity of animal well-being to ensure productivity, while non-use
values include the remaining issues that provide utility to the farmer,
although they are not directly related to productivity or even coun-
tervailing. For decision makers with high non-use values, it is reason-
able to allocate inputs towards animal health and well-being although
such actions neither directly nor indirectly correspond to increased
productivity. The question remains whether such investments in FAW
come at the cost of productivity or may even ensure economic effi-
ciency or both.
Another strand of research has emerged analysing the direct con-
tribution of FAW towards productivity. For instance, Lawson et al.
(2004b) found no relationship between milk production, technical ef-
ficiency and reproductive disorders, as did Hansson and Öhlmér (2008).
Contrary to this, higher technical efficiency could be proven to be as-
sociated with a lower frequency of milk fever but also with a higher rate
of digestive disorders (Lawson et al., 2004a). In addition, Barnes et al.
correlated with higher levels of technical efficiency. As Lawson et al.
(2004a) emphasise, the choice of variables to measure animal welfare
may influence empirical evidence given the complex nature of farm
animal welfare, where health is just one dimension. Resulting re-
lationships between productivity and animal welfare are therefore
H.D. Schulte et al. Land Use Policy 79 (2018) 375–385

method dependent and finding an appropriate measure remains a assessment of farm animal welfare according to the WQP consists of
challenge. several measures used to determine welfare criteria (cf. Table A1 in the
Two general ideas to measure farm animal welfare exist (cf. Appendix). These welfare criteria are summarised by four indicators:
Henningsen et al., 2018). Resource-based measures, such as space in the housing, feeding, behaviour and health each range from 0 to 100,
barn, water availability or playing/brushing facilities, are directly where the overall assessment takes the mean of these four indicators
under the control of the farmer. These offer the main advantage con- (cf. Table A1). Measuring farm animal welfare according to the WQP
cerning predictability and assessment. For instance, policymakers offers the advantage that it directly relates to the animals’ needs and
might increase the required space for each cow and dairy farmers could behavioural patterns. In turn, dairy farmers also benefit from the use of
react by adjusting their management accordingly. Such measures bear the WQP since the index gives a basis for comparison to improve the
the disadvantage of not evaluating the animal’s constitution directly. conditions of animals (Gieseke et al., 2014). According to the scheme of
Outcome-based measures, in contrast, essentially focus on assessing the the WQP, pasture access enters the measure directly. That is, pasture
animals (Botreau et al., 2009) and to a lesser extent on the use of re- would increase the score compared to confinement systems, even if all
sources (Botreau et al., 2013). Even though these measures report a remaining measures are identical. To investigate exactly this relation-
realistic picture of the animal, data collection is time-consuming and ship, we follow Armbrecht et al. (2015) and argue that a higher score
expensive. In our study, we make use of the advantages of outcome- for the majority of the sample farms by system and not by cow-based
based evaluation of dairy cow welfare by relying on the evaluation assessment would not adequately reflect the situation. To ensure com-
scheme of the EU’s Welfare Quality project (cf. Keeling, 2009). How the parability of welfare without such biases, we left out the positive
welfare status of livestock relates to specific husbandry systems, such as scoring by pasture access in the WQP. Data were collected twice per
pasture-based milk production, and whether higher farm animal wel- farm by one person, first at the end of the pasture season (July–October
fare goes along with sacrificing milk productivity if opting for such 2014) and second at the end of the barn season (January–April 2015).
systems, is thus far still under debate. According to Lusk and Norwood We base our calculations on mean values of both visits. The total WQP
(2011), lower productivity would mainly be due to the more extensive index shows comparably high levels for the sampled farms with rather
nature of pasture-based systems but ensuring higher FAW. However, low levels of variation across groups. For greater details we refer to
the dairy farm manager, as caretaker of the animals, influences the Armbrecht et al. (2015) and Table A2 in the Appendix.
production process with his or her knowledge, skills and preferences. We can rely on information on dairy branch-specific costs for the
These might be sequential and a farmer might decide to first achieve economic year 2013/14 and 2014/15 including major branch-specific
efficiency in milk production and then prioritise input allocation to- cost positions such as those for feed, labour, veterinary costs, land,
wards farm animal welfare whilst still maintaining efficiency (cf. depreciation and implicit interest for capital. Given that the WQP refers
Asmild and Hougaard, 2006). Henningsen et al. (2018) even argue that to 2014/15, we concentrate on this year regarding the economic data
the production process determines animal welfare and that given its but present all results for the year 2013/14 as a robustness check in the
complexity, the respective animal welfare status might likewise affect Appendix (see Table A3 for descriptive statistics).
the production process as a prerequisite to achieving productive per- Overall, no dairy farm of our sample consists of more than 280 cows
formance. If in addition, for instances driven by non-use values, FAW and hence the sample farms could be categorised as medium-large
contributes to the farmer’s utility, then FAW likely adds to consumers’ compared to the average across Germany. On average, dairy farms in
utility. Farm animal welfare may even be characterized as a public good Germany have 61 cows and in the state of Lower Saxony, 85.5 cows
provided by farmers since nobody can be excluded from consumption (Destatis, 2017, p. 57), where the group with longest grazing times
(e.g., Bennett, 1995; Lusk, 2011).2 (> 10 h) in the sample operates at a mean herd size of about 121 cows
Our approach addresses these issues by enhancing the output vector (cf. Table 1). The group with medium grazing times (6–10 h) reveals a
by farm animal welfare within a directional distance function approach mean herd size of about 140 cows. Comparing milk yields, those farms
to measure technical efficiency. This allows us to investigate firstly the in the confinement group and the group with < 6 h pasture access show
potential trade-off between FAW and productivity and secondly whe-
ther pasture-based systems have an advantage in the provision of farm Table 1
animal welfare. Descriptive statistics for the year 2014/15.
Pasture Pasture Pasture Confinement
> 10 h 6 h – 10 h <6h
3. Study region and data
Herd size [#] mean 120.58 134.52 140.17 125.10
Our dataset contains data for branch-specific cost and profits for 45 SD 47.44 49.44 64.48 57.84
Milk [kg ECM per cow and day] mean 27.09 26.75 30.60 30.61
dairy farms in the Federal State of Lower Saxony, in north-western
SD 3.23 3.24 2.67 2.92
Germany. These farms are located within one of the main milk pro- Profit [ct. per kg ECM] mean −3.75 −7.20 −6.52 −9.64
ducing regions in Germany, where cows typically have pasture access. SD 3.11 4.15 4.17 4.55
The grazing period usually ranges from April to September. Our dataset WQP overall index [0,100] mean 47.02 47.11 48.36 45.28
includes farms with different levels of pasture length if pasture access SD 5.11 3.47 6.59 3.98

was at least 120 days per year, measured by an indicator: > 10 h, Costs [ct. per kg ECM]
6–10 h, or < 6 h pasture access per day. Besides, the data includes a Labour mean 8.93 9.42 8.90 9.69
SD 1.62 1.43 1.47 1.52
control group with year-round confinement systems, in similar ranges
thereof for pasture and silage mean 3.46 4.02 3.76 4.29
of herd size to ensure comparability, and presupposed willingness to Feed mean 11.83 12.71 12.98 14.00
collaborate. SD 1.26 1.48 1.47 1.24
We rely on the WQP for cattle (Keeling, 2009; Welfare Quality®, thereof for pasture and silage mean 3.64 4.18 3.42 3.62
2012) to assess dairy cow welfare, which has been classified primarily Veterinary mean 1.13 1.25 1.54 1.37
SD 0.81 0.43 0.27 0.29
as a outcome-based measure (Fraser et al., 2013). The overall Land and capital mean 0.96 1.43 1.02 1.16
SD 0.35 0.64 0.32 0.30
N 12 12 13 8
2
This argumentation, however, has been critically debated: some authors
argue that animal welfare is not a public good Mann (2005), while Harvey and Note: Own calculations; SD = standard deviation; N denotes the number of
Hubbard (2013) classify the regulations of farm animal welfare levels as a farms per group out of a total of 45 farm observations.
public good.

377
H.D. Schulte et al.
the highest yields. On average, these cows produce 3 kg of energy-
corrected milk (ECM) per day more than the cows in the groups with
longer pasture access. The sampled farms’ operate above the German
average of 7746 kg per cow per year in 2016, presuming 305 days of
lactation (BLE, 2017). Furthermore, and in comparison to the year
2013/2014 (see Appendix Table A3), the farm profits in 2014/2015
were relatively low. Year 2014/15 was characterized by low milk
prices, which could be partly explained by the Russia embargo in au-
tumn 2014, where important export markets broke away. In addition,
the end of the European Union’s milk quota scheme was in April 2015
and many farms have taken this opportunity to expand production
quantity already before with non-negligible supply effects.
All costs are related to the production of 1 kg of energy-corrected
milk (ECM) with 4.0% fat and 3.4% protein. Labour and feed amount
are the main cost positions and reflect approximately 60% of the costs.
Labour, feed and veterinary costs vary by group: the group with
6 h – 10 h pasture time show the highest labour cost among the pasture
groups, while the confinement group reveals the highest mean labour
cost overall (cf. Table 1). Pasture-based milk production is known to be
labour-intensive, as supported by these numbers, though more hired
labour and potentially higher qualifications might be needed in con-
finement systems. Furthermore, our dataset allows us to differentiate
how the cost of feed and labour were related to the production of silage
and pasture. Related to this differentiation, the labour costs for the
pasture and silage production are slightly higher for confinement sys-
tems than for pasture-based systems. However, similar patterns do not
hold for feed costs.
4. Empirical strategy
The aim of our study is to empirically investigate the relationship
between pasture access, dairy cow welfare based on the scientific farm
animal welfare indicator WQP, and farm performance. We rely on
technical efficiency as a major dimension of farm performance obtained
by a Data Envelopment Analysis (DEA).3 Against the backdrop that
FAW cannot be priced, it remains a challenge to identify the trade-off
between productivity and farm animal welfare in an allocative manner.
Also, a long-lasting discussion exists about whether technical and eco-
nomic efficiency coincide. Hence, we refer to previous studies’ results
that more efficient farms tend to have higher levels of income (Bravo-
Ureta and Rieger, 1991; Dartt et al., 1999). To justify this measure, we
relate the obtained efficiency scores to the profit measure of the milk
production branch of the dairy farms in our sample using a linear re-
gression.
The DEA approach rooting back to Charnes et al. (1978), known as a
benchmarking tool, follows the idea of generating the best-practice
production frontiers and evaluating relative efficiencies of different
entities (Bogetoft and Otto, 2011). A main advantage of the DEA is that
no ex-ante assumptions about the production frontier must be made.
The possibility of including monetary input and output factors together
with non-monetary ones such as the WQP offers another advantage,
particularly in contexts where the time intense data collection limits the
total number of farm observations (cf. Korhonen and Luptacik, 2004;
Pérez Urdiales et al., 2016; Picazo-Tadeo et al., 2012; Sutherland et al.,
2012).
We approach the empirical investigation from two sides. We start
with a standard DEA output-oriented approach, where we consider
farm animal welfare and milk yield as two outputs.4 Besides our interest
to investigate the cow welfare-milk production relationship, at the cow
3
We refer to Lakner and Breustedt (2017) and the cited literature therein
who present an excellent overview of organic farming efficiency, in which dairy
farm performance in particular is addressed.
4
Throughout all steps of the efficiency analysis, we have used the R package
Benchmarking (Bogetoft and Otto, 2015).
378
Land Use Policy 79 (2018) 375–385
level an output-oriented production can be further justified given that
farms under the EU’s milk quota scheme adjust the number of cows to
meet the quantity restriction but optimize milk yields for given input
capacities. Under the standard approach, established by Färe et al.
(1985) referring to Coelli et al. (2005) for further details, the idea is to
measure how the k -th farm could radially expand the output vector y k
as much as possible to become efficient. The boundary is a piece-wise
linear production function representing the technically efficient output.
The radial expansion of the output vector y k produces a projected point
on the surface of this technology. This framework offers a starting point
as typically carried out in the literature, where we implicitly assume
that farm animal welfare and milk output restrict input use in the same
manner. To approach the relevance of FAW as an output, we apply the
variable selection procedure by Pastor et al. (2002). As suggested by
Sirvent et al. (2005), the output may be treated as irrelevant if after
dropping the output farm animal welfare, the TE score of technically
efficient farms changes by more than 10%; this must hold for at least
15% of the farms.
The standard modelling approach, however, may neglect that dairy
farm managers decide upon sequential preferences. That is, it might be
the primary goal to achieve a high productivity (milk yield) first and
then, sequentially farm animal welfare becomes a relevant goal.
Alternatively, it might even be that farms opt for pasture-access because
of beneficiary effects for the animals and accept lower milk yields, al-
though the questions remain whether this comes at the cost of techni-
cally efficiency. From a societal perspective, another question arises,
namely whether pasture-based systems provide farm animal welfare
more efficiently. To address these questions, we switch in a second step
to a directional measurement. Directional DEA refers a proportional
adjustment approach based on directional distance functions
(Chambers et al., 1996). In contrast to the fixed direction of measuring
efficiency, directional distance functions allow us to evaluate the levels
of efficiency in any direction from the observation points (Asmild et al.,
2003; Bogetoft and Hougaard, 1999). This gives us the opportunity to
measure the efficiency related to a specific variable while keeping the
remaining outputs fixed. Interpreting the efficiency score of dairy farms
will either be based on adjusting farm animal welfare ( yFAW ) or milk
yield ( yMI ). Hence, we are able to quantify the improvement potential of
FAW efficiency while keeping the level of milk yield fixed. The direc-
tional output distance function is defined as:
⎯→
⎯
D0 (x , y; gy ) = max{β : (y + βgy ) ∈ P (x )}.
The value of β represents the distance between the observation
y = (yMI , yFAW ) and a point on the production frontier ( y + βgy ). The
direction vector, g = (g yMI , g yFAW ) , determines the direction in which
efficiency is measured. We repeated this procedure twice, (i) in the
direction of milk yield to get technical efficiency, by assuming g = (1,0) ,
and (ii) calculating FAW-efficiency by assuming g = (0,1) . Given the
production technology P (x ) and the direction vector g , the directional
distance function determines the maximum feasible expansion of the
two outputs to reach the production frontier. Hence, given the output-
oriented view, the larger the distance between frontier and observation
is, the higher is the efficiency score. To obtain the output directional
distance β k for the k th farm, we followed the DEA procedure of
Bogetoft and Otto (2011) also within the R package Benchmarking.
While milk yield and FAW determine the two-dimensional output,
we consider labour and feed cost as the main inputs in both steps be-
cause they reflect approximately 60% of the costs. Given the low
number of observations relative to the number of potential inputs, we
again followed the selection procedure to identify additional relevant
inputs among veterinary, land and capital costs. In case of over-speci-
fication we would assess a large portion of the farm observations as
efficient or nearly efficient (Podinovski and Thanassoulis, 2007). As
Asmild and Hougaard (2006), we presume variable returns to scale in
both steps since the convexity assumption under the variable returns to
H.D. Schulte et al.
scale assumption leads to the most cautious estimates of the respective
efficiency scores. To measure if, and how, economies of scale matter,
we carry out a regression-based approach, where herd size is regressed
on the technical and FAW efficiency scores estimated by Ordinary Least
Squares (OLS) as suggested by Hoff (2007) and McDonald (2009). Fi-
nally, to investigate how profit is related to technical and FAW-effi-
ciency, we regressed the respective efficiency scores on the income of
the dairy branch, which is again estimated by OLS.
5. Results and discussion
The variable selection procedure determined land, capital and ve-
terinary costs, as non-relevant inputs besides labour and feed for the
dairy farms in our sample (see Table A4 in the Appendix). Hence, all
DEA steps rest on inputs labour and feed costs as inputs. Starting with
the findings from the standard DEA approach, we consider two outputs,
milk and farm animal welfare. For comparison reasons, we present the
model with one output (milk) as well (cf. Table 2). Since we refer to an
output-oriented approach, farms with a TE-Score higher than 1, all
outputs must be proportionally expanded by the efficiency score to be
technically efficient. Reported scores equal to 1 indicate a fully-efficient
farm operating on the production frontier. Comparing by group, we find
no substantial differences in efficiencies, except for farms in the ex-
tensive pasture group with more than 10 h, which show slightly higher
efficiencies despite the lower level of milk yield. Moreover, some ten-
dency that farms with confinement systems could improve efficiency by
increasing milk yield prevails under both model specifications. Using
the Wilcoxon-rank sum test as suggested by Banker et al. (2010), equal
scores between these two groups are rejected (p-value = 0.0242 DEA-
Base; p-value = 0.0273 DEA FAW-output; other differences are re-
jected). Based upon the selection procedure and given the minor dif-
ferences in mean DEA scores of both models (cf. Table 2), these findings
suggest farm animal welfare to be a non-relevant output and non-de-
terminant of the production frontier in milk production for our sample
(cf. Table A4). These findings may suggest that at first glance dairy
farmers in this study seem to optimize milk yield rather than farm
animal welfare irrespective of pasture-access. When comparing the
mean of the efficiency scores between groups across years, the results
are not similar, yet it still holds that FAW is not a relevant output (see
Table A5 in the Appendix for the efficiency analysis referring to the
year 2013/14).
Dairy farms, however, might rank milk yield and farm animal
welfare differently, where an optimisation towards milk yields might be
the primary goal. Differentiating the dimensions of the output vector by
using a directional DEA allows us to capture whether farms face se-
quential preferences in a sense that improvement potentials concerning
farm animal welfare will be realized after technical efficiency is
achieved. In Table 3, we report the mean efficiency scores in both di-
rections by group, where we refer to “pure” technical efficiency in a
classical sense in the direction of the milk output while the level of FAW
is kept constant. Vice versa, FAW-efficiency refers to efficiency in the
Table 2
Standard DEA results for the year 2014/15.
DEA: milk output DEA: milk and FAW output
Confinement mean 1.23 1.23
SD 0.14 0.14
< 6 h pasture mean 1.15 1.13
SD 0.15 0.14
6 h-10 h pasture mean 1.14 1.12
SD 0.11 0.10
> 10 h pasture Mean 1.08 1.06
SD 0.07 0.06
Note: output-orientation under variable returns to scale; N = 45 farms.
379
Land Use Policy 79 (2018) 375–385
Table 3
Technical and FAW Efficiency (2014/15) - directional distance DEA.
Technical Efficiency FAW-Efficiency
Confinement Mean 1.23 1.53
SD 0.14 0.36
< 6 h pasture Mean 1.16 1.29
SD 0.15 0.34
6 h-10 h pasture Mean 1.15 1.25
SD 0.11 0.18
> 10 h pasture Mean 1.07 1.20
SD 0.06 0.19
Note: output-orientation and variable returns to scale were considered. N = 42
because for three farms no efficiency score could be generated.
direction of animal welfare, that is, the score captures potential in-
creases in FAW to reach the frontier while the milk yield is held con-
stant but for the same input combination (labour and feed cost). FAW-
efficiency can be interpreted in a similar manner as the technical milk
efficiency: a high score for instance, reveals higher efficiency in pro-
viding farm animal welfare. That is, for the same level of feed and input
cost a higher level of farm animal welfare is achieved compared to
farms with a lower FAW-efficiency score.
The results of the directional DEA again reveal some differences
between the mean technical efficiency of the farms in the confinement
system group and the extensive pasture based system keeping the cows
for more than 10 h on the pasture (Wilcoxon rank sum test; p-
value = 0.028). This also holds for the FAW-efficiency direction, where
we find differences between mean scores of confinement and all pas-
ture-based groups. According to the rank-sum test we cannot reject
differences if cows have pasture access with more than 10 h (p-
value = 0.075) and between 6 h and 10 h (p-value = 0.098) against
confinement, respectively. For the year 2013/14, we do find similar
patterns, though not at significant levels (see Table A6 in the Ap-
pendix). From this, we conjecture that pasture-based farms in the
sample operate at higher efficiency in both directions, even though at
lower milk yields on average (cf. Table 1). These findings are somewhat
contrary to those reported by Allendorf and Wettemann (2015), who
did not find any effect of pasture; however, these authors did not
consider farm animal welfare as variable in their DEA and purely rely
on health indicators and a simple dummy variable indicating pasture
access in a second stage regression.
The directional DEA approach as pursued here offers a more so-
phisticated view given the complexity of FAW. While in the standard
DEA model FAW is identified as a non-relevant output, the results of the
directional DEA indicate that farms in the pasture-groups more effi-
ciently provide FAW with a weak tendency that FAW-efficiency may
increase in pasture-access time. For the sample dairy farms pasture-
access seems to be a preferable production option as a result of opti-
mizing farm performance at given conditions. Since these farms do not
benefit from a specific pasture-milk program or any other policy mea-
sures regarding pasture or farm animal welfare, pasture-access can be
treated as a result of farmers’ rationale irrespective of debated ex-
pectations on livestock production systems by society. Even though the
findings suggest that pasture access could even facilitate a better per-
formance of dairy farms in both directions, we cannot not conclude that
pasture-access guarantees higher levels of animal welfare, also because
peer farms that determine the efficient frontier stem from all groups.5
To investigate further the potential trade-off between farm animal
welfare and technical efficiency, which is already questioned by the
mean efficiency scores at the group level, we plot the individual FAW-
efficiency scores against the technical efficiency scores (cf. Fig. 1 and
Fig. A1 in the Appendix for the year 2013/14). If there was a trade-off
5
This further supports the assumption of one technology for the production
process.
H.D. Schulte et al.
Fig. 1. Technical and FAW Efficiency (2014/15) - directional distance DEA.
Note: output-orientation and variable returns to scale were considered.
between farm animal welfare and technical efficiency, we would expect
more observations in the upper-left (high FAW-TE, low TE) or in the
lower-right (low FAW-TE, high TE) regions in Fig. 1. As indicated by
Fig. 1, both efficiency scores reveal a positive and significant correla-
tion with each other (cf. Table A7 in the Appendix) suggesting that farm
animal welfare must not be sacrificed by efficiency and vice versa. In
other words, similar and efficient input combinations enhance technical
efficiency as well as FAW-efficiency. Based on these findings, we
question the existence of a trade-off between farm animal welfare and
technical efficiency. We conjecture that farmers’ value farm animal
welfare in the production process. These findings are in contrast to
Hansson and Öhlmér (2008) who could not find any relationship be-
tween technical efficiency and animal health, though their measures
seem not appropriately capturing the complexity of farm animal wel-
fare.
To investigate the relationship between technical and farm animal
welfare efficiency and scale, we regress herd size on the respective ef-
ficiency scores (see Table A8 in the Appendix). We find that larger herds
are linked to higher efficiency scores. Hence, size can interpreted to be
important to realise efficiencies for our sampled farms. This holds not
only for technical efficiency, but also for farm animal welfare. Note-
worthy, these relationships are based on herd sizes that do not exceed
280 cows. In the range of herd sizes of the sample farms, the finding is
plausible; however, we cannot conclude from these findings on con-
siderably larger herd sizes. Referring to Islam et al. (2015), the high
performance of pasture based milk producing farms may no longer hold
under considerably larger herd sizes. For instance, increasing walking
distances for the cows might lead to lower milk yields, milk losses on
the pasture but also to stress for the animals potentially leading to lower
efficiency scores. The given data set does not allow us to investigate the
likely non-linear relation between efficiency in both directions and herd
sizes.
Finally, since we treat technical efficiency as a major dimension of
farm performance and technically more efficient farms might not ne-
cessarily show a higher overall farm performance, we investigate the
relationship between the contribution of the dairy branch to farm
profits and both efficiency measures. Therefore, we regress the effi-
ciency scores on the branch-specific profitability measure, where the
results indicate a positive relationship (see Table A9 in the Appendix).
That is, increasing efficiency in either of the directions fosters profit-
ability. This result is in line with findings from previous studies ana-
lysing technical and economic efficiency in the dairy sector (e.g., Bravo-
Ureta and Rieger, 1991) and supports the positive relationship between
farm performance and farm animal welfare discussed by Henningsen
et al. (2018).
380
Land Use Policy 79 (2018) 375–385
6. Concluding remarks
This study was motivated by the question of how to warrant will-
ingness to pay for pasture-based milk production because of increases in
farm animal welfare along with sacrificing milk productivity from a
production economics perspective. To investigate the relationship be-
tween pasture access, farm animal welfare and production perfor-
mance, we employ a detailed dataset of dairy branch-specific costs and
profit for 45 dairy farms in north-western Germany. By relying on a
scientific outcome-based farm animal welfare measure according to the
Welfare Quality® Protocol, the dataset offers certain advantages over
similar studies with more observations but more simplistic welfare
measures and farm accounting data that are not branch-specific. We
exploit technical efficiency as a central dimension of economic per-
formance by using Data Envelopment Analysis. Based on a standard
approach, where farm animal welfare is considered as a second output,
we find that pasture-based milk production can be as or more efficient
than confinement systems but farm animal welfare as a non-relevant
output. Efficiency scores in terms of a proportional expansion of both
outputs cannot capture sequential preferences such as improvements
with regards to farm animal welfare will be treated after technical ef-
ficiency is achieved and would hence offer misleading interpretations
regarding farm animal welfare. Instead, we suggest a directional dis-
tance function approach. We first find that farm animal welfare and
technical efficiency are positively correlated and no evidence of a trade-
off exists. Second, the results show higher efficiency levels, both on a
technical level and in terms of animal welfare provision, for pasture
based-systems. That is, higher technical efficiency is observed with a
more efficient provision of animal welfare for those offering the animals
access to pasture, despite lower milk yields on average. The results
further reveal a positive relation of size and efficiency for the sample
farms, where the maximum herd size does not exceed 280. These
findings do not support the argumentation that specialization, and
larger herds are per se more demanding and thus challenge pasture-
based milk production, also because the bottleneck availability of sui-
table grassland. Our sample shows that (moderately) larger herds’
pasture access can be managed efficiently, also regarding farm animal
welfare. However, very large herds may no longer benefit from pasture
access due to potentially negative effects on cows’ well-being due to
long walking distances, which could not be analysed in this study.
We conjecture that pasture-based dairy systems may appear see-
mingly (economically) inefficient in this debate, also because ag-
gregated productivity measures or only milk yield are often taken as a
simplified base. But, pasture access does not automatically lead to an
efficient provision of welfare. Given that we find farms from all groups
determining the most efficient peers approximating the production
frontier, we conjecture that higher milk yield and specialization in
confinement systems do not come at the cost of animal welfare per se.
This result somewhat contradicts common beliefs of some consumers
that it is only the small dairy farms with pasture-access that are able to
provide animal welfare and only by sacrificing performance. As such,
our results add to the discussion initiated by Lusk and Norwood (2011)
that changing the animal husbandry system might have an effect on the
productivity and animal welfare, but it is not guaranteed.
From these results, we conclude that governmental support should
rather be designed to improve the management skills of dairy farmers in
order to increase their efficiency in both directions instead of sup-
porting a specific animal husbandry system or purely pasture access.
This would lead to more technically efficient farming, providing more
efficiently farm animal welfare whilst not discriminating against spe-
cific animal husbandry systems. To increase overall levels of livestock
H.D. Schulte et al. Land Use Policy 79 (2018) 375–385

welfare at the aggregate level, we follow the argumentation of
Kehlbacher et al. (2012) and suggest assessing farm animal welfare
directly rather than solely supporting specific systems. Price premiums
depending on the outcome-based level of animal welfare instead of
whether the animals have pasture access or not offer a higher trans-
parency for consumers and can even be justified given the public-good
character of farm animal welfare. However, the cost-intensity of such
policy programs based on direct livestock welfare measures must be
taken into consideration. On-farm measures such as the Welfare Quality
Protocol® are costly and would require a high administration burden.
New technologies such as automated scanning of animals’ body con-
ditions might offer further possibilities for the future design of such
measures to ease the provision of sufficient levels of animal welfare, as
well as transparency in food production.
Given cows’ preferences and the importance of pastures from an
agri-environmental and sustainability point of view asks for a combined
view of an overall farm animal welfare measurement and specific
production systems with pasture access for the animals in such pro-
grammes. Pasture use may foster maintenance of grassland with its
positive environmental effects by ecosystem services (e.g., biodi-
versity). These aspects also constitute public goods induced by pasture
access (Cooper et al., 2009; Drake, 1992; Lefebvre et al., 2014) and
should be considered in policy design. After all, pasture access remains
highly appreciated by consumers, not only in terms of animal welfare,
but also for the aesthetic landscape it provides. The background of the
declining number of grazing animals in Europe supports the argu-
mentation of taking both, farm animal welfare and the role of pastures
within the ecosystem into account. Future research must consider the
environmental aspect of different pasture-based systems explicitly to
inform the policy debate. Due to a lack of data, these dimensions were
not part of our analysis.
Our study must be viewed as a first step towards creating a frame-
work for quantitative animal welfare production economics and offers a
promising path of future research for analysing the production of dif-
ferently ranked goods from a production economics perspective.
Nonetheless, this study has a number of limitations: the low number of
observations limits generalizations to be drawn. Further, the production
process as well as the welfare of the livestock highly depends on
management and personal norms, beliefs and values. Within this study,
we cannot identify to which extent the difference in the technical ef-
ficiency scores can be causally explained by pasture access or man-
agement skills. Future research could consider these issues while in-
vestigating economic efficiency of farms. In addition, given that pasture
access might contribute to robustness and overall lifetime of the ani-
mals, which in turn could be an investment for farms, a dynamic per-
spective would be desirable.
Declaration of interest
All authors confirm that no actual or potential conflict of interest
including any, financial, personal or other relationships with other
people or organizations within three years of beginning the submitted
work exist that could inappropriately influence, or be perceived to in-
fluence, the outcome of the underlying article.
Acknowledgements
The authors gratefully acknowledge financial support from the
Ministry for Science and Culture of Lower Saxony (MWK) within the
collaborative research project SAM, Analysis of Dairy Production:
Grazing versus Indoor Housing of Dairy Cows, Support Code: ZN 286,
University of Rostock and the German Federal Ministry of Food and
Agriculture within the programme for ecological farming
(Bundesprogramm Ökologischer Landbau) for ÖKOTAWEK, Support
Code: 2812NA009.

Appendix A

A.1 Data details

Table A1
Structure of the Welfare Quality® Assessment protocol for cattle (Botreau et al., 2009; Welfare
Quality®, 2012).
Welfare criteria Welfare principle

Absence of prolonged hunger (body condition score) Feeding

Absence of prolonged thirst (functioning of water points)
Comfort around resting (time needed to lie down) Housing
Thermal comfort (as yet no measure is developed)
Ease of movement (presence of tethering)
Absence of injuries (integument alternations) Health
Absence of disease (nasal discharge)
Absence of pain induced by management procedures (tail docking)
Expression of social behaviours (agonistic behaviours) Behaviour
Expression of other behaviours (access to pasture)
Good human-animal relationship (avoidance distance)
Positive emotional state (qualitative behaviour assessment)

Note: Measurement examples for the criteria are given in parentheses. The principles are determined
by the criteria and scores range from 0-100. The overall assessment is the average of the principles.
Access to pasture was not taken into account as a measure.

381
H.D. Schulte et al. Land Use Policy 79 (2018) 375–385

Table A2
Descriptive statistics of the WQP and its principals.
Pasture Pasture 6 h- Pasture Confinement
> 10 h 10 h <6h

Feeding mean 42.55 43.32 53.61 54.15

SD 20.47 17.09 7.10 12.43
Behaviour mean 36.98 34.77 35.09 39.67
SD 3.93 3.85 6.59 3.66
Housing mean 63.37 62.14 61.46 49.02
SD 8.83 6.90 7.73 5.50
Health mean 45.18 48.22 43.29 38.28
SD 7.72 6.86 4.12 8.10
Overall mean 47.02 47.11 48.36 45.28
SD 5.11 3.47 64.48 3.98

Note: for details see Armbrecht et al. (2015).

Table A3
Descriptive statistics for the year 2013/2014.
Pasture Pasture Pasture Confinement
> 10 h 6 h-10 h <6h

Herd size [#] mean 112.16 125.25 134.92 121.74

SD 38.17 49.40 58.36 58.03
Milk yield [kg ECM per cow and day] mean 27.98 28.10 31.19 31.39
SD 2.63 3.16 2.81 1.27
Profit [ct. per kg ECM] mean 2.34 3.78 1.84 4.64
SD 5.02 2.72 4.04 4.55
WQP overall index [0.100] mean 47.02 47.11 48.36 45.28
SD 5.11 3.47 58.36 3.98

Costs [ct. per kg ECM]

Labour mean 10.72 9.87 10.67 10.42
SD 2.28 2.53 1.72 1.84
thereof for pasture and silage mean 3.52 3.88 3.93 3.41
Feed mean 12.65 12.70 13.81 13.42
SD 2.07 1.24 1.48 2.01
thereof for pasture and silage mean 3.12 3.43 3.47 2.71
Veterinary mean 1.12 1.34 1.44 1.26
SD 0.67 0.43 0.44 0.31
Land and capital mean 1.12 1.24 0.99 0.94
SD 0.57 0.49 0.35 0.29
N 12 12 13 8

Note: Own calculations; SD = standard deviation; N denotes the number of farms per group out of a total of 45 farm observations.

A.2 Additional results

Table A4
Variable selection procedure.
Candidate Number of observations of DMUs with TE > 1.10

2013/14 2014/15

Veterinary costs (input) 2 2

Costs for land and capital (input) 1 1
FAW-WQP (output, based on WQP score) 1 1

Note: Variables were considered relevant for the TE of DEA applications if more than 15% of the DMU (more than 6 farms in
our sample) changed their TE score by at least 10% after the decision making unit became efficient and the candidate was
excluded. Note, within all steps evaluating the respective candidates, standard DEA is conducted assuming inputs labour and
feed costs and milk yield as output.

382
H.D. Schulte et al. Land Use Policy 79 (2018) 375–385

Table A5
DEA-base output oriented approach for the year 2013/14.
DEA-Base DEA FAW-output

confinement mean 1.19 1.19

SD 0.20 0.20
< 6 h pasture mean 1.23 1.22
SD 0.15 0.15
6 h-10 h pasture mean 1.16 1.14
SD 0.10 0.11
> 10 h pasture mean 1.14 1.14
SD 0.14 0.14

Note: output-orientation and variable returns to scale were considered.

Table A6
Technical and FAW Efficiency (2013/14) - directional distance DEA and output-oriented.
Technical Efficiency FAW-Efficiency

Confinement mean 1.19 1.46

SD 0.12 0.42
< 6 h pasture mean 1.22 1.46
SD 0.16 0.27
6 h-10 h pasture mean 1.16 1.28
SD 0.11 0.21
> 10 h pasture mean 1.15 1.33
SD 0.14 0.24

Note: output-orientation and variable returns to scale were considered. N = 43 because for two farms no efficiency scores could be
generated.

Fig. A1. Technical and FAW Efficiency (2013/14) - directional distance DEA and output-oriented.

Table A7
Correlation between Technical and FAW Efficiency.
Year Method FAW-Efficiency

Technical Efficiency 2013/14 Kendall 0.78***

Spearman 0.59***
2014/15 Kendall 0.84***
Spearman 0.64***

Note:*** indicates significantly rejecting the null hypothesis for zero coefficients according to t-test at the 1% level.

Table A8
Size effects on efficiency (OLS).
Dependent variable Year Constant Size Coefficient

FAW-Efficiency 2013/14 1.5941*** −0.0018*

2014/15 1.6460*** −0.0026***
Technical Efficiency 2013/14 1.2951*** −0.0009**
2014/15 1.2998*** −0.0012***

Note: OLS regression for efficiencyi = α 0 + βherd herdsizei + εi , where εi denotes the error term. ***,** denotes the respective significance level to reject the null
hypotheses of βherd = 0 and α 0 = 0 of a t-test at the 1% and 5% level, respectively.

383
H.D. Schulte et al.
Table A9
Influence of Technical and FAW-Efficiency on farm profit.
Dependent variable Year
Profit 2013/14
2014/15
Note: OLS regression for : profiti = α 0 + βeff efficiencyi + εi , where εi denotes the error term. ***,** denotes the respective significance level to reject the null hypotheses
of βeff = 0 and α 0 = 0 of a t-test at the 1% and 5% level, respectively.
References
Allendorf, J.J., Wettemann, P.J.C., 2015. Does animal welfare influence dairy farm effi-
ciency? A two-stage approach. J. Dairy Sci. 98 (11), 7730–7740. https://doi.org/10.
3168/jds.2015-9390.
Armbrecht, L., Lambertz, C., Albers, D., Gauly, M., 2015. Tierwohl von Milchkühen bei
Stall-und Weidehaltung—Ein Vergleich anhand des Welfare Quality Protokolls. In:
1st ed. In: Busch, G., Gieseke, D., Ikinger, C., Kühl, S., Pirsich, W. (Eds.), Tierhaltung
im Spannungsfeld von Tierwohl, Ökonomie und Gesellschaft. Tierwohl-Tagung in
Göttingen, 07.- 08. Oktober 2015, vol. 1. Georg-August-Universität Department für
Agrarökonomie und Rurale Entwicklung, Göttingen, pp. 70–72.
Arnott, G., Ferris, C.P., O’Connell, N.E., 2017. Review: Welfare of dairy cows in con-
tinuously housed and pasture-based production systems. Animal 11 (2), 261–273.
https://doi.org/10.1017/S1751731116001336.
Asmild, M., Hougaard, J.L., 2006. Economic versus environmental improvement poten-
tials of Danish pig farms. Agric. Econ. 35 (2), 171–181. https://doi.org/10.1111/j.
1574-0862.2006.00150.x.
Asmild, M., Hougaard, J.L., Kronborg, D., Kvist, H.K., 2003. Measuring inefficiency via
potential improvements. J. Prod. Anal. 19 (1), 59–76. https://doi.org/10.1023/
A:1021822103696.
Baird, L.G., O’Connell, N.E., McCoy, M.A., Keady, T.W.J., Kilpatrick, D.J., 2009. Effects of
breed and production system on lameness parameters in dairy cattle. J. Dairy Sci. 92
(5), 2174–2182. https://doi.org/10.3168/jds.2008-1333.
Banker, R.D., Zheng, Z., Natarajan, R., 2010. DEA-based hypothesis tests for comparing
two groups of decision making units. Eur. J. Oper. Res. 206 (1), 231–238. https://doi.
org/10.1016/j.ejor.2010.01.027.
Barker, Z.E., Leach, K.A., Whay, H.R., Bell, N.J., Main, D.C.J., 2010. Assessment of la-
meness prevalence and associated risk factors in dairy herds in England and Wales. J.
Dairy Sci. 93 (3), 932–941. https://doi.org/10.3168/jds.2009-2309.
Barnes, A.P., Rutherford, K.M.D., Langford, F.M., Haskell, M.J., 2011. The effect of la-
meness prevalence on technical efficiency at the dairy farm level: an adjusted data
envelopment analysis approach. J. Dairy Sci. 94 (11), 5449–5457. https://doi.org/
10.3168/jds.2011-4262.
Bennema, S.C., Ducheyne, E., Vercruysse, J., Claerebout, E., Hendrickx, G., Charlier, J.,
2011. Relative importance of management, meteorological and environmental fac-
tors in the spatial distribution of Fasciola hepatica in dairy cattle in a temperate
climate zone. Int. J. Parasitol. 41 (2), 225–233. https://doi.org/10.1016/j.ijpara.
2010.09.003.
Bennett, R., 1995. The value of farm animal welfare. J. Agric. Econ. 46 (1), 46–60.
https://doi.org/10.1111/j.1477-9552.1995.tb00751.x.
BLE, 2017. Milchleistung je Kuh in Deutschland in den Jahren 1900 bis 2016 (in
Kilogramm). Accessed 28 November 2017. https://de.statista.com/statistik/daten/
studie/153061/umfrage/durchschnittlicher-milchertrag-je-kuh-in-deutschland-seit-
2000/.
Bogetoft, P., Hougaard, J.L., 1999. Efficiency evaluations based on potential (non-pro-
portional) improvements. J. Prod. Anal. 12 (3), 233–247. https://doi.org/10.1023/
A:1007848222681.
Bogetoft, P., Otto, L., 2011. Benchmarking With DEA, SFA, and R. Springer New York,
New York, NY 351 pp.
Bogetoft, P., Otto, L., 2015. Package’ Benchmarking’. Accessed 30 April 2018. https://
cran.r-project.org/web/packages/Benchmarking/Benchmarking.pdf.
Botreau, R., Veissier, I., Perny, P., 2009. Overall assessment of animal welfare: strategy
adopted in welfare quality. Anim. Welf. 18 (4), 363–370.
Botreau, R., Winckler, C., Velarde, A., Dalmau, A., Butterworth, A., Keeling, L.J., Veissier,
I., 2013. Integration of data collected on farms or at slaughter to generate an overall
assessment of animal welfare. In: Blokhuis, H., Miele, M., Veissier, I., Jones, B. (Eds.),
Improving Farm Animal Welfare. Science and Society Working Together: the Welfare
Quality Approach. Wageningen Academic Publishers, Wageningen, pp. 147–174.
Bravo-Ureta, B.E., Rieger, L., 1991. Dairy farm efficiency measurement using stochastic
frontiers and neoclassical duality. Am. J. Agric. Econ. 73 (2), 421. https://doi.org/10.
2307/1242726.
Breustedt, G., Latacz-Lohmann, U., Tiedemann, T., 2011. Organic or conventional?: op-
timal dairy farming technology under the EU milk quota system and organic sub-
sidies. Food Policy 36 (2), 223–229. https://doi.org/10.1016/j.foodpol.2010.11.019.
Burow, E., Thomsen, P.T., Sørensen, J.T., Rousing, T., 2011. The effect of grazing on cow
mortality in Danish dairy herds. Prev. Vet. Med. 100 (3-4), 237–241. https://doi.org/
10.1016/j.prevetmed.2011.04.001.
Burow, E., Thomsen, P.T., Rousing, T., Sørensen, J.T., 2013. Daily grazing time as a risk
factor for alterations at the hock joint integument in dairy cows. Animal 7 (1),
Land Use Policy 79 (2018) 375–385
Efficiency Constant Coefficient
FAW 4.10 −0.98
Technical −14.07*** −9.58**
FAW 0.38 −4.78**
Technical −13.49** −17.54***
160–166. https://doi.org/10.1017/S1751731112001395.
Cardoso, C.S., Hotzel, M.J., Weary, D.M., Robbins, J.A., von Keyserlingk, M.A.G., 2016.
Imagining the ideal dairy farm. J. Dairy Sci. 99 (2), 1663–1671. https://doi.org/10.
3168/jds.2015-9925.
Chambers, R.G., Chung, Y., Färe, R., 1996. Benefit and distance functions. J. Econ. Theory
70 (2), 407–419. https://doi.org/10.1006/jeth.1996.0096.
Chapinal, N., Goldhawk, C., Passillé, A.M., de, Keyserlingk, M.A.G., von, Weary, D.M.,
Rushen, J., 2010. Overnight access to pasture does not reduce milk production or
feed intake in dairy cattle. Livest. Sci. 129 (1-3), 104–110. https://doi.org/10.1016/j.
livsci.2010.01.011.
Chapinal, N., Barrientos, A.K., von Keyserlingk, M.A.G., Galo, E., Weary, D.M., 2013.
Herd-level risk factors for lameness in freestall farms in the northeastern United
States and California. J. Dairy Sci. 96 (1), 318–328. https://doi.org/10.3168/jds.
2012-5940.
Charlton, G.L., Rutter, S.M., East, M., Sinclair, L.A., 2011. Preference of dairy cows: in-
door cubicle housing with access to a total mixed ration vs. access to pasture. Appl.
Anim. Behav. Sci. 130 (1-2), 1–9. https://doi.org/10.1016/j.applanim.2010.11.018.
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making
units. Eur. J. Oper. Res. 2 (6), 429–444. https://doi.org/10.1016/0377-2217(78)
90138-8.
Coelli, T.J., Prasada Rao, D.S., O’Donnell, C.J., Batesse, G.E., 2005. An Introduction to
Efficiency and Productivity Analysis, 2nd ed. Springer, New York XVII, 349 S.
Cooper, T., Hart, K., Baldock, D., 2009. Provision of public goods through agriculture in
the European Union. Report Prepared for DG Agricultural and Rural Developement,
Contract No 30-CE-02330911/00-28. Institute for European Enviromental Policiy,
London 396 pp.
Dartt, B.A., Lloyd, J.W., Radke, B.R., Black, J.R., Kaneene, J.B., 1999. A comparison of
profitability and economic efficiencies between management-intensive grazing and
conventionally managed dairies in Michigan. J. Dairy Sci. 82 (11), 2412–2420.
https://doi.org/10.3168/jds.S0022-0302(99)75492-5.
de Vries, M., Bokkers, E.A.M., van Reenen, C.G., Engel, B., van Schaik, G., Dijkstra, T., de
Boer, I.J.M., 2015. Housing and management factors associated with indicators of
dairy cattle welfare. Prev. Vet. Med. 118 (1), 80–92. https://doi.org/10.1016/j.
prevetmed.2014.11.016.
Destatis, 2017. Land- Und Forstwirtschaft, Fischerei: Viehbestand Und Tierische
Erzeugung. Fachserie 3 Reihe 4, Wiesbaden, 2030400167004.
Drake, L., 1992. The non-market value of the Swedish agricultural landscape. Eur. Rev.
Agric. Econ. 19 (3), 351–364. https://doi.org/10.1093/erae/19.3.351.
Ellis, K.A., Billington, K., McNeil, B., McKeegan, D.E.F., 2009. Public opinion on UK milk
marketing and dairy cow welfare. Anim. Welf. 18 (3), 267–282.
Färe, R., Grosskopf, S., Lovell, C.A.K., 1985. The Measurement of Efficiency of
Production. Springer, Dordrecht 216 pp.
Fraser, D., Duncan, I.J.H., Edwards, S.A., Grandin, T., Gregory, N.G., Guyonnet, V.,
Hemsworth, P.H., Huertas, S.M., Huzzey, J.M., Mellor, D.J., Mench, J.A., Spinka, M.,
Whay, H.R., 2013. General principles for the welfare of animals in production sys-
tems: the underlying science and its application. Vet. J. (Lond., Engl.: 1997) 198 (1),
19–27. https://doi.org/10.1016/j.tvjl.2013.06.028.
Gieseke, D., Lambertz, C., Traulsen, I., Krieter, J., Gauly, M., 2014. Beurteilung von
Tiergerechtheit in der Milchviehhaltung – Evaluierung des Welfare Quality
Protokolls. Züchtungskunde 86 (1), 58–70.
Gocsik, É., Brooshooft, S.D., Jong, I.C., de, Saatkamp, H.W., 2016. Cost-efficiency of
animal welfare in broiler production systems: a pilot study using the Welfare Quality®
assessment protocol. Agric. Syst. 146, 55–69. https://doi.org/10.1016/j.agsy.2016.
04.001.
Hansson, H., Lagerkvist, C.J., 2015. Identifying use and non-use values of animal welfare:
evidence from Swedish dairy agriculture. Food Policy 50, 35–42. https://doi.org/10.
1016/j.foodpol.2014.10.012.
Hansson, H., Öhlmér, B., 2008. The effect of operational managerial practices on eco-
nomic, technical and allocative efficiency at Swedish dairy farms. Livest. Sci. 118 (1-
2), 34–43. https://doi.org/10.1016/j.livsci.2008.01.013.
Harvey, D., Hubbard, C., 2013. Reconsidering the political economy of farm animal
welfare: an anatomy of market failure. Food Policy 38, 105–114. https://doi.org/10.
1016/j.foodpol.2012.11.006.
Henningsen, A., Czekaj, T.G., Forkman, B., Lund, M., Nielsen, A.S., 2018. The relationship
between animal welfare and economic performance at farm level: a quantitative
study of Danish pig producers. J. Agric. Econ. 69 (1), 142–162. https://doi.org/10.
1111/1477-9552.12228.
Hoff, A., 2007. Second stage DEA: comparison of approaches for modelling the DEA score.
Eur. J. Oper. Res. 181 (1), 425–435. https://doi.org/10.1016/j.ejor.2006.05.019.
Ingenbleek, P.T.M., Immink, V.M., Spoolder, H.A.M., Bokma, M.H., Keeling, L.J., 2012.
384
H.D. Schulte et al.
EU animal welfare policy: developing a comprehensive policy framework. Food
Policy 37 (6), 690–699. https://doi.org/10.1016/j.foodpol.2012.07.001.
Islam, M.R., Garcia, S.C., Clark, C.E.F., Kerrisk, K.L., 2015. Modelling pasture-based au-
tomatic milking system herds: system fitness of grazeable home-grown forages, land
areas and walking distances. Asian Australas. J. Anim. Sci. 28 (6), 903–910. https://
doi.org/10.5713/ajas.14.0385.
Keeling, L. (Ed.), 2009. An Overview of the Development of the Welfare Quality®
Assessment Systems. Univ. School of City and Regional Planning, Cardiff V, 97 S.
Kehlbacher, A., Bennett, R., Balcombe, K., 2012. Measuring the consumer benefits of
improving farm animal welfare to inform welfare labelling. Food Policy 37 (6),
627–633. https://doi.org/10.1016/j.foodpol.2012.07.002.
Korhonen, P.J., Luptacik, M., 2004. Eco-efficiency analysis of power plants: an extension
of data envelopment analysis. Eur. J. Oper. Res. 154 (2), 437–446. https://doi.org/
10.1016/S0377-2217(03)00180-2.
Lagerkvist, C.J., Hess, S., 2011. A meta-analysis of consumer willingness to pay for farm
animal welfare. Eur. Rev. Agric. Econ. 38 (1), 55–78. https://doi.org/10.1093/erae/
jbq043.
Lakner, S., Breustedt, G., 2017. Efficiency analysis of organic farming systems‐a review of
concepts, topics, results and conclusions. German J. Agric. Econ. 66 (2), 85–108.
Lawson, L.G., Agger, J.F., Lund, M., Coelli, T., 2004a. Lameness, metabolic and digestive
disorders, and technical efficiency in Danish dairy herds: a stochastic frontier pro-
duction function approach. Livest. Prod. Sci. 91 (1-2), 157–172. https://doi.org/10.
1016/j.livprodsci.2004.07.016.
Lawson, L.G., Bruun, J., Coelli, T., Agger, J.F., Lund, M., 2004b. Relationships of effi-
ciency to reproductive disorders in Danish milk production: a stochastic frontier
analysis. J. Dairy Sci. 87 (1), 212–224. https://doi.org/10.3168/jds.S0022-0302(04)
73160-4.
Lefebvre, M., Espinosa, M., Gomez y Paloma, S., Paracchini, M.L., Piorr, A., Zasada, I.,
2014. agricultural landscapes as multi-scale public good and the role of the common
agricultural policy. J. Environ. Plann. Manage. 58 (12), 2088–2112. https://doi.org/
10.1080/09640568.2014.891975.
Legrand, A.L., Keyserlingk, M.A.G., von, Weary, D.M., 2009. Preference and usage of
pasture versus free-stall housing by lactating dairy cattle. J. Dairy Sci. 92 (8),
3651–3658. https://doi.org/10.3168/jds.2008-1733.
Lusk, J.L., 2011. The market for animal welfare. Agric. Hum. Values 28 (4), 561–575.
https://doi.org/10.1007/s10460-011-9318-x.
Lusk, J.L., Norwood, F.B., 2011. Animal welfare economics. Appl. Econ. Perspect. Policy
33 (4), 463–483. https://doi.org/10.1093/aepp/ppr036.
Mann, S., 2005. Ethological farm programs and the “market” for animal welfare. J. Agric.
Environ. Ethics 18 (4), 369–382. https://doi.org/10.1007/s10806-005-7049-y.
McDonald, J., 2009. Using least squares and tobit in second stage DEA efficiency analyses.
Eur. J. Oper. Res. 197 (2), 792–798. https://doi.org/10.1016/j.ejor.2008.07.039.
McInerney, J., 2004. Animal Welfare, Economics and Policy: Report on a Study
Undertaken for the Farm & Animal Health Economics Division of Defra. DEFRA,
Exeter, UK Accessed 15 November 2016. http://webarchive.nationalarchives.gov.
uk/20110318142209/http://www.defra.gov.uk/evidence/economics/foodfarm/
reports/documents/animalwelfare.pdf.
Motupalli, P.R., Sinclair, L.A., Charlton, G.L., Bleach, E.C., Rutter, S.M., 2014. Preference
and behavior of lactating dairy cows given free access to pasture at two herbage
masses and two distances. J. Anim. Sci. 92 (11), 5175–5184. https://doi.org/10.
2527/jas.2014-8046.
Napolitano, F., Pacelli, C., Girolami, A., Braghieri, A., 2008. Effect of information about
animal welfare on consumer willingness to pay for yogurt. J. Dairy Sci. 91 (3),
910–917. https://doi.org/10.3168/jds.2007-0709.
Pastor, J.T., Ruiz, J.L., Sirvent, I., 2002. A statistical test for nested radial dea models.
385
Land Use Policy 79 (2018) 375–385
Oper. Res. 50 (4), 728–735. https://doi.org/10.1287/opre.50.4.728.2866.
Pérez Urdiales, M., Lansink, A.O., Wall, A., 2016. Eco-efficiency among dairy farmers: the
importance of socio-economic characteristics and farmer attitudes. Environ. Resour.
Econ. 64 (4), 559–574. https://doi.org/10.1007/s10640-015-9885-1.
Picazo-Tadeo, A.J., Beltrán-Esteve, M., Gómez-Limón, J.A., 2012. Assessing eco-efficiency
with directional distance functions. Eur. J. Oper. Res. 220 (3), 798–809. https://doi.
org/10.1016/j.ejor.2012.02.025.
Pieralli, S., Hüttel, S., Odening, M., 2017. Abandonment of milk production under un-
certainty and inefficiency: the case of western German farms. Eur. Rev. Agric. Econ.
44 (3), 425–454. https://doi.org/10.1093/erae/jbx001.
Podinovski, V.V., Thanassoulis, E., 2007. Improving discrimination in data envelopment
analysis: some practical suggestions. J. Prod. Anal. 28 (1-2), 117–126. https://doi.
org/10.1007/s11123-007-0042-x.
Reijs, J.W., Daatselaar, C.H.G., Helming, J.F.M., Jager, J., Beldman, A.C.G., 2013.
Grazing Dairy Cows in North-West Europe: Economic Farm Performance and Future
Developments With Emphasis on the Dutch Situation. LEI Wageningen UR, The
Hague 124 pp.
Roosen, J., Dahlhausen, J.L., Petershammer, S., 2016. Acceptance of animal husbandry
practices: the consumer perspective. Proceedings in Food System Dynamics, 2016:
Proceedings in System Dynamics and Innovation in Food Networks 2016 260–267.
https://doi.org/10.18461/pfsd.2016.1630.
Schreiner, J.A., Hess, S., 2017. The role of non-use values in dairy farmers’ willingness to
accept a Farm animal welfare programme. J. Agric. Econ. 68 (2), 553–578. https://
doi.org/10.1111/1477-9552.12203.
Schuppli, C.A., Keyserlingk, M.A.G., von, Weary, D.M., 2014. Access to pasture for dairy
cows: responses from an online engagement. J. Anim. Sci. 92 (11), 5185–5192.
https://doi.org/10.2527/jas.2014-7725.
Sirvent, I., Ruiz, J.L., Borrás, F., Pastor, J.T., 2005. A Monte Carlo evaluation of several
tests for the selection of variables in DEA models. Int. J. Info. Tech. Dec. Mak. 4 (03),
325–343.
Sutherland, L.-A., Gabriel, D., Hathaway-Jenkins, L., Pascual, U., Schmutz, U., Rigby, D.,
Godwin, R., Sait, S.M., Sakrabani, R., Kunin, W.E., Benton, T.G., Stagl, S., 2012. The
‘Neighbourhood Effect’: a multidisciplinary assessment of the case for farmer co-or-
dination in agri-environmental programmes. Land Use Policy 29 (3), 502–512.
https://doi.org/10.1016/j.landusepol.2011.09.003.
Vance, E.R., Ferris, C.P., Elliott, C.T., McGettrick, S.A., Kilpatrick, D.J., 2012. Food intake,
milk production, and tissue changes of Holstein-Friesian and Jersey x Holstein-
Friesian dairy cows within a medium-input grazing system and a high-input total
confinement system. J. Dairy Sci. 95 (3), 1527–1544. https://doi.org/10.3168/jds.
2011-4410.
Vanderstichel, R., Dohoo, I., Sanchez, J., Conboy, G., 2012. Effects of farm management
practices and environmental factors on bulk tank milk antibodies against gastro-
intestinal nematodes in dairy farms across Canada. Prev. Vet. Med. 104 (1-2), 53–64.
https://doi.org/10.1016/j.prevetmed.2011.09.022.
Veissier, I., Butterworth, A., Bock, B., Roe, E., 2008. European approaches to ensure good
animal welfare. Appl. Anim. Behav. Sci. 113 (4), 279–297. https://doi.org/10.1016/
j.applanim.2008.01.008.
Weinrich, R., Kuehl, S., Zuehlsdorf, A., Spiller, A., 2014. Consumer attitudes in Germany
towards different dairy housing systems and their implications for the marketing of
pasture raised milk. Int. Food Agribus. Manage. Assoc. 17 (4), 205–221.
Welfare Quality®, 2012. Welfare Quality® Assessment Protocol for Cattle. Welfare Quality
Consortium, Lelystad, Netherlands 182 pp.
Wettemann, P.J.C., Latacz-Lohmann, U., 2017. An efficiency-based concept to assess
potential cost and greenhouse gas savings on German dairy farms. Agric. Syst. 152,
27–37. https://doi.org/10.1016/j.agsy.2016.11.010.