363

© 2009 Universities Federation for Animal Welfare Animal Welfare 2009, 18: 363-370
The Old School, Brewhouse Hill, Wheathampstead, ISSN 0962-7286
Hertfordshire AL4 8AN, UK

Overall assessment of animal welfare: strategy adopted in

Welfare Quality®
R Botreau*†, I Veissier† and P Perny‡
†
INRA, UR1213 Herbivores, F-63122 Saint-Genès-Champanelle, France
‡
LIP6, University of Paris 6, 104 Avenue du Président Kennedy, F-75016, Paris, France
* Contact for correspondence and requests for reprints: [email protected]

Abstract

Animal welfare is multidimensional; its assessment relies on complementary measures covering all dimensions. Welfare Quality®
constructed a multicriteria evaluation model for its assessment at unit level (farms, slaughterhouses). Four welfare principles are
distinguished (‘Good feeding’, ‘Good housing’, ‘Good health’, and ‘Appropriate behaviour’). An animal unit receives four principle
scores (expressed on a 0–100 value scale). These scores are aggregated together to form the overall assessment by sorting animal
units into predefined welfare categories boundaried by reference profiles. A unit is assigned to the welfare category above the
profile it is considered at least as good as. Several assignment procedures were tested on a set of 69 dairy farms and compared
with observers’ general impressions. The welfare categories, reference profiles and assignment procedure were defined in consul-
tation with social scientists, animal scientists and stakeholders. Four welfare categories were defined: ‘Excellent’, ‘Enhanced’,
‘Acceptable’, and ‘Not classified’. The reference profiles were set at 80, 55 and 20, corresponding to aspiration values for
Excellent, Enhanced and Acceptable. The assignment procedure resulted from a compromise between theoretical opinion on what
should be considered excellent, enhanced or acceptable, and what can realistically be achieved in practice: to be assigned to a
given category, a unit must reach its aspiration value on 2 or 3 of the 4 principles, and not score below the aspiration value for
the next lowest category on the other principle(s). The model can be used for several purposes, including identifying welfare
problems on a farm to advise farmers, or checking compliance with certification schemes.

Keywords: advice, animal welfare, certification, expert opinion, multicriteria evaluation, overall assessment

Introduction Europe and between stakeholders, it is essential that this

The aims of the European Welfare Quality® (2004–2009) overall evaluation is based on a formal model (see Botreau
project were to design methods for the overall assessment of et al 2007a). The general architecture of the evaluation
cattle, pig, and poultry welfare, on-farm and at slaughter. model developed in Welfare Quality® was outlined in
This assessment tool is designed to help farmers and slaugh- Botreau et al (2007b). Briefly, Welfare Quality® identified
terhouse managers identify welfare problems and monitor 12 key animal welfare criteria: absence of prolonged
progress. It can also provide information to consumers on hunger; absence of prolonged thirst; comfort around
the welfare of the animals from which products are sold. resting; thermal comfort; ease of movement; absence of
Welfare is a multidimensional concept that requires all the injuries; absence of disease; absence of pain induced by
component dimensions to be checked by specific indicators management procedures; expression of social behaviour;
(eg Mason & Mendl 1993; Fraser 2003; Botreau et al expression of other behaviours; good human-animal rela-
2007a). Therefore, the Welfare Quality® assessment incor- tionship and absence of general fear (Botreau et al 2007).
porates numerous welfare measures focused essentially on Data collected on an animal unit are used to check unit
animals and to a lesser extent on resources or animal compliance with the 12 welfare criteria. The scores obtained
management, whether on-farm or in slaughterhouses. For at criterion level are then collated to assess unit compliance
the purposes of this paper, we will use the word unit to refer with four main welfare principles (Good feeding, Good
to either a farm or a slaughterhouse. The Welfare Quality® housing, Good health, and Appropriate behaviour). Finally,
assessment generates a substantial amount of data that these principle scores are used to conclude on an overall
needs be interpreted in terms of welfare and then integrated evaluation (Table 1). The model was fine-tuned in response
to produce an overall evaluation at unit level. For the to expert views from a panel of animal scientists, social
assessment to be used routinely and consistently across scientists, and stakeholders.

Universities Federation for Animal Welfare Science in the Service of Animal Welfare
364 Botreau et al

Table 1 Sequential evaluation structure, from measures collected on-farm through the welfare criteria and principles
and, finally, the overall assessment (Botreau et al 2007c). Criteria and principles are common to all animal types under
study (pigs, cattle, and poultry). Measures are specific to animal types. The table includes the measures designed for
dairy cows, on-farm (Botreau et al 2008).

Measurement in dairy cows Welfare criteria (12) Welfare principles (4)

Body condition score Absence of prolonged hunger
Good
Provision of water (drinker [n] and cleanliness) Absence of prolonged thirst feeding
Behaviour around resting, cleanliness Comfort around resting
No reliable measure available at present Thermal comfort Good
Presence of tethering, access to exercise Ease of movement housing Overall
Skin alteration, lameness Absence of injuries Overall
assessment
Respiratory, digestive, reproductive diseases Absence of disease Good
health
Dehorning, tail docking (rare) Absence of pain induced by management procedures
Incidence of aggressive behaviours Expression of social behaviours
Access to pasture Expression of other behaviours Appropriate
Avoidance distance to human approach Good human-animal relationship behaviour
Qualitative assessment of behaviour Positive emotional state

The definition and construction of the criteria and principles weights the different criteria while strongly limiting
were described in previous papers (Botreau et al 2007c; between-criteria compensations.
Botreau et al 2008) and will be only summarised here as
From principles to the overall assessment: what problem to
background information. The present article focuses on the
answer?
final stage of the evaluation model, ie producing the overall
assessment from the four principle scores (expressed on a To produce an overall assessment at unit level, we have to
0–100 value scale). aggregate the four principle scores into a single evaluation.
The choice of aggregation method depends closely on the
Background information type of issue the stakeholders intend to resolve with an
overall assessment, ie the type of decision they want to be
Initial steps in the development of the welfare assessment
able to take based on the unit evaluations. For example, if
model: from measures to principles
the aim is to select the best unit, an optimisation method
The 12 welfare criteria proposed by Welfare Quality® may be relevant, but this is not the case if the aim is to
were discussed in 55 consumer-citizen focus groups distinguish several categories of welfare offered to animals.
across seven European countries. They received wide- Decision-aiding systems commonly define four main
spread agreement, and consumer-citizens rejected the idea reference problems (Roy 1996):
that some criteria may be more important than others
• the description of units based on a number of unit charac-
(Veissier & Evans 2007).
teristics, expressed in an appropriate language shared by all
Welfare Quality® partners checked the relevance, repeata- stakeholders;
bility and feasibility of numerous welfare measures and
• the choice of a subset of units, that is kept as narrow as
finally defined 30 to 60 measures on animal units for each
possible in order to facilitate the final choice of only one
animal type studied (dairy cows, fattening bulls, veal calves,
unit, ie the best on offer;
broilers, layers, sows and piglets, and fattening pigs; see
Veissier et al [2007] for a description of the measures). • the sorting of units, achieved by assigning each unit into
Animal unit compliance with each welfare criterion identi- pre-defined categories defined in terms of norms (ie situa-
fied in the project is checked by one or several measures. At tions that stakeholders consider as typifying the different
this stage, data produced by relevant measures on each categories);
criterion are turned into welfare value scores on a common • the ranking of units, from best to worst.
scale ranging from 0 to 100, where 0 corresponds to the The description is a prerequisite for the three other types of
worst situations, 100 to ideal situations, and 50 is for problems, because the description has to be formalised to
‘neither good nor bad’ situations. Any animal unit scoring make it possible to issue recommendations (which is the
below 20 on any single criterion has a high risk of being objective of the three other types of problems). However,
excluded from any certification scheme that includes the description alone is not sufficient to generate a recom-
welfare concerns. The 12 criterion scores are integrated to mendation which inevitably includes value judgements. The
form 4 scores at principle level using an operator that Welfare Quality® assessment tool has been developed as a

© 2009 Universities Federation for Animal Welfare

 Overall assessment of animal welfare 365

decision aid for stakeholders, ie to help farmers design each unit to one of the defined welfare categories. The
welfare improvement strategies, to help certification bodies Management Committee took the final decision on the
decide whether or not to incorporate units into a certifica- reference profiles. Concerning the assignment process,
tion scheme, to help consumers decide whether or not to several of the procedures put forward by the task force were
buy a certain product, etc. Therefore, the assessment model tested on a set of real farms. We thus assessed 69 randomly
shall result in neither the choice of best unit nor in the chosen dairy farms in Austria, Germany and Italy with a
ranking of all units, but rather in an absolute evaluation of view to obtaining farms in each welfare category, so as to
the level of welfare offered by the units to the animals, with check that the welfare assessment was able to detect differ-
welfare defined in several categories, from very low to ences between farms. The observers visiting farms in
excellent welfare. Consequently, in the context of Welfare Austria and Germany (n = 44) also gave their own general
Quality®, the overall assessment of animal welfare corre- impression of the farms they visited, using a 12-cm visual-
sponds to a sorting problem aimed at assigning animal units analogue scale, from a very bad to an excellent level of
to predefined ordered categories of welfare. welfare. We ran logistic regressions to check whether the
farm classifications obtained using the different procedures
Materials and methods
fitted with the general impression of the observers auditing
The animal units are to be assigned to welfare categories the farms. A Likelihood-Ratio test was used to estimate the
according to the scores they obtained on the four welfare strength of these links between the observer’s impression of
principles (Good feeding, Good housing, Good health, a given farm and the category reached by that farm
Appropriate behaviour) expressed on the 0–100 value according to each of the proposed procedures. Finally, we
scale. We decided to use a multicriteria analysis method- consulted the Advisory Committee during a second group
ology similar to ELECTRE TRI (see, for example, meeting on the most appropriate procedure to be applied,
Moscarola & Roy 1977) in order to respect the multidi- taking into account theoretical aspects, the distribution of
mensionality of animal welfare (Mousseau & Slowinski the 69 tested farms, and the agreement of the resulting farm
1998). Briefly, the animal units are compared to reference categorisation system with observers’ general impressions.
profiles outlining the categories to be distinguished. The
tool is then developed through three steps: (i) definition Results
of the welfare categories; (ii) position of the reference
profiles and (iii) definition of the rules for comparing Definition of welfare categories
units with reference profiles. Four scenarios were defined for overall welfare assessment
To go through these three steps, we consulted both scientists tool utilisation. These scenarios are not mutually exclusive:
(in social and animal sciences) and stakeholder groups with Scenario 1
an interest in the development of an animal welfare assess- The assessment system is used to set a minimum level of
ment tool. Several groups of experts were thus consulted: animal welfare that is acceptable; it could either replace the
the Welfare Quality® Advisory Committee composed of current EU regulations on minimum animal welfare
representatives from a producer group, a retailer, two standards or help in implementing a regulation on compul-
animal protection groups, the Federation of Veterinarians of sory labelling with a yes/no label or a system using several
Europe, and political institutions (the World Organisation levels (similar to the one used for energy consumption by
for Animal Health and the European Commission household appliances).
Directorate General for ‘Health and Consumers’), a task
force composed of 5 social scientists and 9 animal scien- Scenario 2
tists, all partners of Welfare Quality®, and the Management The assessment system is used to distinguish instances in
Committee of the project, made up of 2 social scientists and which the welfare of the animals not only exceeds the
6 animal scientists. minimum acceptable level but attains high levels, sufficient
An overview of the welfare assessment method was to implement a voluntary labelling system.
presented at the first meeting between the Management Scenario 3
Committee and the Advisory Committee. Together with the The assessment system is used by the unit manager as a self-
Management Committee, we suggested a list of potential assessment management tool to identify welfare problems
uses stakeholders could make of an overall animal welfare or risks, and to monitor improvements.
assessment tool. We then asked the individual Advisory
Scenario 4
Committee members to react to and comment on the
proposed scenarios, specifying their own organisations’ The assessment system is used as a research tool to evaluate
views on potential uses of the Welfare Quality® assessment farming systems and practices.
system. The task force collated these views to establish a In order to be implementable, each of these four scenarios
final set of scenarios representing all the potential uses at requires a number of welfare categories to be distinguished.
least one organisation considered interesting, defined The number of categories required may differ between
welfare categories so as to implement these scenarios, scenarios: from two categories for Scenario 1 to at least four
proposed profiles to boundary welfare categories, and categories for Scenarios 3 and 4 (Table 2). This means that
initiated discussion on possible procedures for allocating all four of the following categories are needed to enable

Animal Welfare 2009, 18: 363-370

366 Botreau et al

Table 2 Implications of potential uses on the welfare categories distinguished.

Scenario Potential uses of the welfare assessment Welfare categories needed

n Description
1 Standard for cross-compliance and future definition of a 2 Below/average minimum legal requirements or
minimum animal welfare standard equivalent (if no legislation)
Compulsory labelling defining several levels of welfare 4 Poor/Normal/Good/Excellent
2 Setting welfare targets for farm certification schemes: 3 Very high welfare/High welfare/Other
Voluntary labelling
3 Feedback to producers for monitoring the results of
welfare improvement strategies Very poor-----------------------------------------Very high
Several
Self assessment tool level of welfare (with intermediate categories)
4 Assessing new animal farming systems/breeds
Several Very poor-----------------------------------------Very high
Furthering research on animal welfare level of welfare (with intermediate categories)

Table 3 Utilisation of welfare categories for each potential scenario. Depending on the use targeted, the response could
be binary (eg is the level of welfare in accordance with legal requirements?) or gradual (is the farm –, +, ++ or +++).

Welfare categories Voluntary label (Niche Voluntary label (Quality Compulsory label Self-assessment or research
market: scenario 2) label: scenario 2) (scenario 1) tool (scenarios 3 and 4)
Excellent Yes Yes Yes +++
Enhanced No Yes Yes ++
Acceptable No No Yes +
Not classified No No No –

simultaneous implementation of all four scenarios cited reference profile corresponds to a set of four values
above: expressed on the 0–100 scale, one per principle.
• Excellent, the welfare of the animals is excellent — the The 0–100 scale used to express principle-scores has the
animal unit may correspond to a niche market, via a label same meaning whatever the principle. Therefore, we chose
guaranteeing consumers very high quality products (this to set flat reference profiles, ie identical scores on all prin-
label could be dedicated to animal welfare); ciples (Figure 1). This also appeared consistent with the
• Enhanced, the welfare of the animals is good but not idea that all welfare dimensions are equally important (see
excellent — good farming practices are applied and are Background information). The experts consulted to
sufficient to ensure a good level of welfare within a more construct the principles were told that at below 20, a farm
general quality label which could encompass other product would not be considered acceptable. Therefore, 20 was
characteristics, like food quality, food safety, environmental chosen for Profile 3 (ie the lower limit for the ‘Acceptable’
impact; category). The experts were also told that 50 means ‘neither
• Acceptable, the welfare of the animals is acceptable good nor bad’. Hence, a slighter higher score (55) was
(above minimal requirements defined for a compulsory chosen for Profile 2 (lower limit for the ‘Enhanced’
label), but insufficient for the farm to enter a certification category). Profile 1 (lower limit of the ‘Excellent’ category)
scheme based on respect for animal welfare; and was set at 80 to be symmetrical to Profile 1. The four scores
constituting a given profile correspond to ‘aspiration
• Not classified, the welfare of the animals is low and
values’ to be achieved by an animal unit to reach the
considered unacceptable.
category set by this profile.
The use of the above welfare categories would differ from
one scenario to another, leading to either a binary answer or Rules for comparing farms with reference profiles
a graded answer (Table 3).
Elaboration of a set of potential procedures
Reference profile The scores obtained by an animal unit on the four welfare
Since four ordered welfare categories are to be distin- principles are to be compared to the scores for profiles 1, 2
guished, three reference profiles have to be defined corre- and 3. Rules need to be set to decide when an animal unit is
sponding to the limits between consecutive categories. Each considered at least as good as a profile. An intuitive rule is

© 2009 Universities Federation for Animal Welfare

 Overall assessment of animal welfare 367

Figure 1

Reference profiles boundarying welfare

categories. A Farm x is compared to each
profile and then assigned to the welfare
category above the profile that the farm
is considered at least as good as.

unanimity, whereby a unit is considered better than a profile requirement of that category on all principles, and reach the
if it scores higher than that profile on all welfare principles, aspiration value on 3 of them, eg to be considered excellent,
ie it always reaches the aspiration values set for that an animal unit must score at least 55 (Profile 2, minimum
category. If a unanimity rule is taken, then, to be excellent, requirement) on all principles and reach 80 (Profile 1, aspi-
an animal unit needs to score at least 80 on every principle, ration value) on three of them.
to be considered enhanced it needs to score at least 55 on Procedure 3, First set of mixed rules (2/3/4):
every principle, and to be considered acceptable it needs to
score at least 20 on every principle. Less strict rules may be • to be considered excellent, an animal unit must score at
applied where, to be assigned to a given category, a unit least 55 (Profile 2) on all principles and reach 80 (Profile 1)
only has to score better than the aspiration values of that on 2 of them;
category on a given number of principles. However, it • to be considered enhanced, an animal unit must score at
would be inappropriate to consider a unit as excellent if it least 20 (Profile 3) on all principles and reach 55 (Profile 2)
scores more than 80 on three principles but only 10 on the on 3 of them;
fourth one. Consequently, even when more lenient rules are • to be considered acceptable, an animal unit must score at
applied, there is a fixed minimum requirement for the least 20 (Profile 3) on all 4 principles;
remaining principles. In our assessment system, this
• otherwise, the animal unit is not classified.
minimum generally equals the aspiration values of the next
lowest category. For instance, we may decide to consider an Procedure 4, Second set of mixed rules (2/2/3):
animal unit as excellent if it scores higher than 80 on two • to be considered excellent, an animal unit must score at
principles (80 being the aspiration value for excellence) and least 55 (Profile 2) on all principles and reach 80 (Profile 1)
higher than 55 on the two remaining principles (55 being on 2 of them;
both the minimum requirement for excellence and the aspi- • to be considered enhanced, an animal unit must score at
ration value for the enhanced category). This kind of rule least 20 (Profile 3) on all principles and reach 55 (Profile 2)
system makes it necessary to define a Profile 4 correspon- on 2 of them;
ding to the minimum requirement for the bottom-but-one
• to be considered acceptable, an animal unit must score at
category (acceptable). We decided to set Profile 4 at 10 for
least 10 (Profile 4) on all principles and reach 20 (Profile 3)
all principles, corresponding to half the scores of Profile 3.
on 3 of them;
We thus investigated five different procedures for assigning
animal units to welfare categories (see below): • otherwise, the animal unit is not classified.
Procedure 1, Unanimity rule (4/4/4): to be assigned to a Procedure 5, Third set of mixed rules (1/2/3):
given category, an animal unit needs to reach the aspiration • to be considered excellent, an animal unit must score at
value of that category on all 4 principles, eg to be consid- least 55 (Profile 2) on all principles and reach 80 (Profile 1)
ered excellent, an animal unit must score at least on 1 of them;
80 (Profile 1) on all principles. • to be considered enhanced, an animal unit must score at
Procedure 2, Qualified-majority rule (3/3/3): to be assigned least 20 (Profile 3) on all principles and reach 55 (Profile 2)
to a given category, an animal unit must reach the minimum on 2 of them;

Animal Welfare 2009, 18: 363-370

368 Botreau et al

Table 4 Consequences of the different procedures on how five virtual farms are categorised, the farms are charac-
terised by their scores on the four principles, noted here P1, P2, P3 and P4 for Good feeding, Good housing, Good
health and Appropriate behaviour, respectively. The five farms are differentiated on P2.

Principle scores Welfare category according to each procedure*

Farm number P1 P2 P3 P4 Procedure 1 Procedure 2 Procedure 3 Procedure 4 Procedure 5
1 60 90 95 70 Enhanced Enhanced Excellent Excellent Excellent
2 60 65 95 70 Enhanced Enhanced Enhanced Enhanced Excellent
3 60 45 95 70 Acceptable Enhanced Enhanced Enhanced Enhanced
4 60 14 95 70 Not classified Acceptable Not classified Acceptable Acceptable
5 60 4 95 70 Not classified Not classified Not classified Not classified Not classified

*See text for explanation on procedures.

• to be considered acceptable, an animal unit must score at Procedures 1, 2 and 3 led to non-significant links with the
least 10 (Profile 4) on all principles and reach 20 (Profile 3) observers’ impressions (χ2 = 2.58, 1.28 and 3.80, P > 0.05).
on 3 of them; The five procedures were presented to the Welfare
• otherwise, the animal unit is not classified. Quality® Advisory Committee, together with the farm
In all cases, the sorting procedure follows a descending order: distributions and the data on concordance with observers’
for a given farm, we first check whether it can be assigned to general impressions. Procedures 1 and 3 were excluded
the category ‘Excellent’, and if not, we check whether it can be because of being either too severe (allowing few or no
assigned to the second-best category ‘Enhanced’, and so on. farms to reach the ‘Enhanced’ category) or not reflecting
To illustrate the differences between Procedures 1 to 5, we observers’ impressions. Procedure 2 was excluded because
tested them on a set of five virtual farms (Table 4). These it was not discriminatory between farms (the vast majority
farms were characterised by the same principle-scores on of farms were ‘Acceptable’) and not reflecting observers’
P1, P3 and P4 (ie respectively, Good feeding, Good housing impressions. Procedure 5 appeared too flexible, as only
and Appropriate behaviour), but differed on P2 (Good one score out of four needed be excellent for a farm to be
housing). Table 4 clearly illustrates how the five procedures considered as ‘Excellent’ all-round. Procedure 4 was
differ in severity (to the extent of a difference of one finally retained as it appeared to offer a good compromise
category for a given farm), the most severe being Procedure between severity, discrimination of farms, and concor-
1 (unanimity) and the most indulgent being Procedure 5. dance with observers’ general impressions.
Choice of the most appropriate procedure Discussion and conclusions
Figure 2 shows the distribution of the 69 dairy farms The construction of the assessment system proposed in
visited across the four welfare categories according to Welfare Quality® respects the multidimensionality nature of
the five assignment procedures described above. Using animal welfare, relying on four main principles to be fulfilled
Procedures 1 and 3, 51% of the 69 farms fell in the to ensure welfare. In addition, the assessment model allows
lowest category (not classified), no farms or only 3% the formalisation of the reasoning followed by animal scien-
were enhanced, and no farm was deemed excellent. tists, social scientists and potential users for scoring the
Procedure 2 clearly led to a non-discriminatory distribu- animal units. The model relies on both theoretical aspects (eg
tion, with 83% of farms in the ‘Acceptable’ category. for the definition of the reference profiles boundarying the
With Procedures 4 and 5, a majority of the farms (67%) four welfare categories) and on what can be realistically
fell in the category ‘Acceptable’ and 14% of farms were achieved in practice (eg data collected within Welfare
not classified. Still no farms were considered excellent. Quality® on real farms were used to choose appropriate
After a close look at the data collected on these farms, it assignment procedures). The following procedure is
proved reasonable to consider that none of the farms proposed for assigning animal units to welfare categories
should be considered excellent. Indeed, there was a high based on the scores obtained on the four principles: a farm is
incidence of lameness, injuries and disease, and no farm considered ‘excellent’ if it scores more than 55 on all princi-
could be considered excellent on two welfare principles. ples and more than 80 on two of them; it is considered
Procedures 4 and 5 revealed a significant relationship ‘enhanced’ if it scores more than 20 on all principles and
between the welfare categories and the general impression more than 55 on two of them; it is considered ‘acceptable’ if
of the observers who audited the farms, with χ2 values of it scores more than 10 on all principles and more than 20 on
5.56 and associated probabilities of 0.02, respectively. three of them. If the farm fails the final condition, it is deemed

© 2009 Universities Federation for Animal Welfare

 Overall assessment of animal welfare 369

not classified. This procedure still needs to be tested on the Figure 2

other animal types studied in Welfare Quality®.
Since the system formalises the reasoning of experts while
describing all the calculations necessary to go from
measures to overall assessment, it can easily be standard-
ised and automated for routine use. This is an essential
characteristic for an evaluation system to be used within
certification schemes or other systems that require a
normative procedure. To be implementable, this kind of
assessment system has to be transparent, with information
accessible to any interested party. To ensure standardisa-
tion and availability of information, technical documents
are currently prepared to ensure such transparency with
the help of the Nederlands Normalisatie-instituut (NEN).
In parallel, a software tool is being developed to facilitate
on-farm or in-slaughterhouse data collection and storage,
and streamline the automated calculation of welfare scores
and categorisation of animal units.
We are aware that the proposed assessment system is not
perfect. Any method relying on the definition of a small
number of categories will always carry threshold effects,
and a farm or slaughterhouse manager may not have an
interest in first improving the most problematic point
detected on their farm/slaughterhouse. This could be the
case of a farm with 35 on good feeding, 55 on good housing,
20 on good health and 50 on appropriate behaviour, which
would be categorised as acceptable. The most problematic
point of this farm is health, yet improving its health score
from 20 to 54, which would be a major improvement, would
still not be enough to upgrade its classification. On the
contrary, to become enhanced, the farmer could opt to work
on increasing its behaviour score just 5 points from 50 to 55,
which would only be a relatively small improvement.
Hence, there may be only limited benefits for the animals.
However, this non-desired effect is minimised by the fact
that it is risky for the farmer to remain at a score around
20 on health, because a slip of just one point on this
principle would be sufficient to make the farm drop into the
‘not classified’ category.
To be fully efficient in helping farmers and slaughterhouse
managers improve the level of animal welfare, the assessment
system proposed in Welfare Quality® should be completed by
advice provided by a person qualified to identify the root
causes of the problems highlighted through the assessment
process. Once these causes have been properly detected, the
unit manager can be offered remedial solutions. The assess-
ment system described may well help advisers to identify
causes of poor welfare by focusing attention on specific
issues. It may thus act as a useful step towards improving the
global level of welfare offered to animals in Europe
throughout their life, from farm to slaughter.
Animal welfare implications
The assessment system elaborated by Welfare Quality®
(including the description of measures and the calculation of
scores) could be seen as a reference method serving several Distributions of 69 European dairy farms between the four wel-
purposes. The assessment can be used: i) to provide farmers fare categories, using the five proposed procedures (Figures 3a,
with a broad picture of the welfare status of their animals and 3b, 3c, 3d, 3e for procedures 1, 2, 3, 4, 5, respectively).

Animal Welfare 2009, 18: 363-370

370 Botreau et al

to identify the aspects requiring attention; ii) to give policy- Botreau R, Veissier I, Butterworth A, Bracke MBM and
makers a better understanding of the welfare implications of Keeling LJ 2007c Definition of criteria for overall assessment of
farming systems and practices; iii) to certify animal units on animal welfare. Animal Welfare 16: 225-228
Botreau R, Capdeville J, Perny P and Veissier I 2008
welfare grounds, by implementing the assessment system on
Multicriteria evaluation of animal welfare at farm level: an applica-
units to be certified or by certifying the system and the tion of MCDA methodologies. Foundations of Computing and
practices employed on these units, and iv) to facilitate Decision Sciences 33: 1-18
informed decision by stakeholders, including consumers. Fraser D 2003 Assessing animal welfare at the farm and
group level: the interplay of science and values. Animal
Acknowledgements
Welfare 12: 433-443
We thank the Welfare Quality® partners and members of Mason G and Mendl M 1993 Why is there no simple way of
the project’s Advisory Committee who acted as experts for measuring animal welfare? Animal Welfare 2: 301-319
fine-tuning the evaluation model. We are also grateful to the Moscarola J and Roy B 1977 Procédure automatique d’ex-
Welfare Quality® partners who visited the dairy farms and amen de dossiers fondée sur une segmentation tri-
provided us with their data. The present study is part of the chotomique en présence de critères multiples. RAIRO
Welfare Quality® research project co-financed by the Recherche Opérationnelle 11: 145-173. [Title translation:
European Commission within the 6th Framework Automatic file examination based on trichotomic segmenta-
Programme, contract No. FOOD-CT-2004-506508. The text tion in presence of multiple criteria]
represents the authors’ views and does not necessarily Mousseau V and Slowinski R 1998 Inferring an ELECTRE TRI
represent a position of the Commission who will not be model from assignment examples. Journal of Global Optimization
liable for the use made of such information. 12: 157-174
Roy B 1996 Multicriteria Methodology for Decision Aiding p 57.
References Kluwer Academic: Dordrecht, The Netherlands
Botreau R, Bonde M, Butterworth A, Perny P, Bracke Veissier I and Evans A 2007 Rationale behind the Welfare
MBM, Capdeville J and Veissier I 2007a Aggregation of meas- Quality® assessment of animal welfare. Proceedings of the 2nd
ures to produce an overall assessment of animal welfare: Part 1, Welfare Quality® Stakeholder Conference pp 19-22. 3-4 May 2007,
A review of existing methods. Animal 1: 1179-1187 Berlin, Germany
Botreau R, Bracke MBM, Perny P, Butterworth A, Veissier I, Forkman B and Jones B 2007 Proceedings of the
Capdeville J, van Reenen CG and Veissier I 2007b Second Welfare Quality® Stakeholder Conference on Assuring Animal
Aggregation of measures to produce an overall assessment of ani- Welfare: From Societal Concerns to Implementation pp 70-84. 3-4
mal welfare: Part 2, Analysis of constraints. Animal 1: 1188-1197 May 2007, Berlin, Germany

© 2009 Universities Federation for Animal Welfare