Consumer preferences for pork supply

chain attributes

M.P.M. Meuwissen 1 ,*, I.A. Van Der Lans 2 and R.B.M. Huirne 1

I Institute for Risk Management in Agriculture, Business Economics Group, Wageningen University,
P.O. Box 8130, NL-6700 EW Wageningen, The Netherlands
2 Marketing and Consumer Behaviour Group, Wageningen University, Wageningen, The Netherlands
* Corresponding author (tel: +31-317-483857; fax: +31-317-482745; e-mail: [email protected])

Received 4 August 2006; accepted 13 January 2007

Abstract
Based on an extensive customized conjoint analysis with 24 attributes of pork production, covering
issues from feed to fork, we identified six consumer segments: ecologists (17%), tradition-minded consumers
(17%), animal friends (16%), health-concerned consumers (18%), economists (12%) and unpronounced
consumers (20%). Typically all segments prefer pork originating from the Netherlands and a zero risk
of Salmondla. Discriminating items between segments include issues of pig breeding, housing, farm-level
handling of pigs, safety aspects such as residue levels and irradiation of pork, and taste and price.
Segments were furthermore found to differ on issues such as willingness to pay for pork production
improvements and pork label perception. From our findings we recommend decision-makers in pork
supply chains to no longer market pork as a bulk product as there are distinct requirements for pork
and the way it is produced. Also, there seems to be sufficient financial room to invest in each of these
segments. At the same time, however, it should be noticed that consumers have difficulties with
distinguishing between different labels and that they generally have limited knowledge about basic
pork production and safety issues.

Additional keywords: market segmentation, customized conjoint analysis, mixture regression models,
willingness to pay, pork labels

Introduction
Producing food in chains and networks started mainly from a business economics
perspective, amongst other to reduce transaction costs (Williamson, 1985; Boehlje &
Lins, 1998). Encouraged by quality assurance schemes, tracking and tracing systems,
and increasing requirements for transparency, the 'feed to fork' principle is now
standard in many regulations. For instance, the European General Food Law (Anon.,
2002) explicitly addresses the traceability of food (ingredients) throughout the entire

NJAS 54-3, 2007 293

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

supply chain. However, in contrast to 'business to business transparency', the commu-

nication of production issues to consumers still is very limited. Consumer information
is generally restricted to country of origin and country of packing (Meuwissen et a!',
zo03). Specific labels, for instance from organically produced products, provide somewhat
more information to consumers, but market shares of such products generally are very
small (Yussefi, zo06).
This also holds for pork in the Netherlands. The majority of fresh pork is retailer-
branded or labelled as Integrated Chain Control (IKB) and the combined market share
of special labels such as EKO (organic), Free range and Biodynamic (also organic)
is less than 1% (Anon., zo06). On the basis of consumer studies that closely mimic
actual pork buying behaviour, Meuwissen & Van Der Lans (zooS) concluded that
consumers generally base their buying decisions on product characteristics such as
taste and price. On the other hand, there have also been studies pointing to the increasing
relevance of production attributes (see amongst other ones Verbeke, ZOOI). SO we
hypothesize that, despite current bulk markets, there are (groups of) consumers who
are interested in pork production issues and greater 'business to consumer transparency'.
Within this framework, our paper aims at (I) identifying consumer segments, based
on numerous pork production attributes covering the entire supply chain; and (z)
describing these segments amongst other things with respect to current consumption
oflabelled pork, perception about existing pork labels, knowledge about pork production,
and willingness to pay for production improvements. We shall focus on pork production
in the Netherlands, which is characterized by very intensive, mostly indoor, farming,
very high pig densities, high export figures (in the Netherlands the degree of self-
sufficiency for pork is z30%) and much public attention for animal welfare, food
safety and environmental issues.
Previous consumer studies about production attributes of meat focus only on a
limited number of aspects, or discuss these aspects qualitatively or at an aggregate
level. For instance, Nayga (1996) studied consumers' perceptions of hormone use,
antibiotics, irradiation and feed from grain produced using herbicides. Den Ouden et
a!. (1997) focused on perceptions about animal welfare issues only. Hoffmann (zooo)
studied perceptions about the country of origin of fresh meat. Also Dransfield et a!.
(zooS) considered the country of origin, complemented by information on indoor or
outdoor raising of the animals. Ngapo et a!. (zo03) qualitatively discussed consumer
perceptions of pork production through focus group discussions. Aggregated concepts
such as 'animal friendly', 'environmentally friendly' and 'safety' were studied by
Schifferstein et a!. (1998) and Verbeke & Viaene (1999)' Krystallis & Arvanitoyannis
(zo06) considered detailed aspects of food safety (Salmonella, dioxin, hormones,
antibiotics) but addressed consumers' interest for 'production method information'
at an aggregate level.
In contrast to these studies, our study incorporates Z4 attributes of pork production,
including 6z attribute levels, referring to issues from feed to fork and addressing
consumer concerns such as food safety, animal welfare and sensory quality. Meuwissen
& Van Der Lans (zooS) reported on general concerns about pork production and the
overall ranking of the Z4 attributes, whereas the current paper focuses on the details
of the attribute levels and market segmentation.

294 NJAS 54-3, 2007

 Consumer preferences for pork attributes

Materials and methods

Consumer data were gathered using a computerized questionnaire. In addition to
introductory questions and questions on socio-economic and (pork) consumption
characteristics, the questions covered four themes: (I) preferences for pork-production
attributes, (2) knowledge of pork production, (3) willingness to pay for pork-production
improvements, and (4) perception of currently existing pork labels. Throughout the
questionnaire we furthermore addressed a number of societal issues related to pork
production. Below are discussed the elicitation of preferences and willingness to pay,
the way the data were collected and the segmentation method to cope with the large
amount of pork-production attributes. The questions about knowledge, perception and
citizen issues are described along with the answers.

Elicitation of consumer preferences

For eliciting consumer preferences we used the customized conjoint analysis (CCA)
as described by Srinivasan & Park (1997)' The CCA analysis included 24 attributes,
which together with their levels are listed in Figure I, grouped under 'feed and breed',
'farm' and 'processing and retail'. There are 22 attributes on the pork-production
process and 2 product attributes: taste and price. The attributes had two, three or four
levels, as shown in Figure I (the remainder of the figure is explained in the Results
section). In developing the computerized CCA analysis, the procedure as described
by Hensel-Borner & Sattler (2000) was used as a basis. Some modifications were imple-
mented, as described below.
The CCA analysis consists of three parts. The respondents were first asked to
give self-explicated desirability ratings for the 22 production attributes. Attributes were
categorized under different topics, i.e., pig feeding, breeding, farming, processing and
retail. The desirability of the levels was rated per attribute on a scale from 0 (least desirable
level) to 10 (most desirable level). That is, respondents had first to select the level they
liked most and the level they liked least. The most desirable level was assigned a score
of 10 on a 0-10 rating scale, the least desirable level a score of o. Next, respondents
were asked to rate the other levels on the same scale. In case of an attribute with only
two levels, respondents had to select the level they liked most. After the self-explicated
desirability ratings were given, respondents were asked to give self-explicated importance
ratings, i.e., to rate the importance of the difference between the most and the least
desirable level for each of the subsequent production attributes on a scale from 0 (not
important at all) to 100 (very important). The midpoint of this scale (50) was labelled
'reasonably important'. Next, as a small check, respondents were asked whether they
thought any important pork production attributes had been left out. Ten per cent of
the respondents replied affirmatively, in spite of the fact that most of the issues that
were indicated had been included in the questionnaire already.
In the third part of the CCA analysis, respondents continued with graded paired
comparisons in which they had to indicate their preference for one partial pork profile
description over another one on a scale from I (strong preference for the one profile)
to 9 (strong preference for the other profile). These partial profiles were described for

NJAS 54-3, 2007 295

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

1lO
Feed&brMd
-<0
€ 20
i
"
~
-20

-40

-60 e a-
! 1;- ll8 o
~ ~
5
~ ~
> f
~
ij!! f
Z

,] a.. III

Residues of hetblc. GM Sllbstances Bonemeal Residuals of numan food

In pig feed In pig feed In pig feed Industty In pig feed Broedr.g goa. GM In breeding

60

-<0

j20
~ Ot--+-'y~--if~t-.---;-
....-.-Ir--.-.-\\---ifL---*---,.*,'--31P;r--""'1~~""
-20
-40

-60
·i a 1 ~
.. i i;; !0 :I ~'§ .~~
~ -g :: 0
z
. 0 ."
" ~
0
~ z ~ z z
31 E
>-
~ 5 cD ~ ::Ii ~ .b
:; ~
~
Q

.e
0
031 on
.,. S
~~
~
III

""
tJ
~ i E
'2!
'f
.
l'
(J
a.. 0
~ .Ii III
;;;
co
~ Sl
'-v--' '-v--' '-v--' '-v--' ' - v - - ' "'----==~
Application of E""'-ntaJ
medicines InsJdelootsJdo Space Housing LMng surface TIBdoddng TeeIl1 cipping Castration requiremenlS

Pro••sslng & retail

.e
~ 20
~
:.ol-...:Itr--~I--\-:lL.._--llr----"';'.....::It----~---J~--+---!';;.p.--~~

·20

:!: i
~
~ :::>
w
§j
E
J! ." .l!

... "" !l
~ .0
.II .II
:t::Ii
J1
.~
§
~
!!
~
§
~
!!
~
..
~ :z:
0 ..
~ :z:
0

... ...8
!!
!!
§ ~
0 III
w
" w
0
.1:
z~ w
:::> 0
.S
~
!!
oS 0 S
:z: ~ ~
~
::Ii Q. III III
~
-----.....---.. -----.....---..
0
----....---- ----....---- ----....---- ----....---- '-.---'
CholaIln Slim""".. R..Id.... Extra coo«Jng 11T1d1l1lon
Tuto Cauntty of origin Traceability quality & pIioe 01 medicines lor solety fo< sorely

- + - EcoJogI.t~ Trad lion-minded - . . - Animal friend ~ Health-concemed ~ Economilt --+- Unpronounced

Figure 1. Pork production attribute-level part-worth estimates per segment.

296 NJAS 54-3, 2007

 Consumer preferences for pork attributes

seven attributes: the two product attributes price and taste, four of the most important
production attributes and one moderately important production attribute. If the production
attributes could not be uniquely determined due to equal importance ratings, then a
random selection was drawn from the equally rated most important and moderately
important attributes, respectively. Pairs of partial profiles were based on a fractional-
factorial main-effects design (Hair et a!., 1998) in combination with a cyclic design
(Huber & Zwerina, 1996). Inadmissible combinations like 'pork originating from the
Netherlands' and 'no traceability', were excluded beforehand.
The CCA procedure mainly differs from the computerized procedure of Hensel-
Borner & Sattler (2000) on two important aspects. Firstly, we did not ask for single-
stimulus partial profile evaluations but for graded paired comparisons. Making such
comparisons more closely resembles the actual shopping situation, and explicitly
confronts respondents with different levels of production attributes. Secondly, the
attributes taste and price were always included in the conjoint analysis, but were
excluded from the self-explicated parts. In our questionnaire design we decided to
do so because otherwise the generally high importance of taste and price might have
masked subtle differences between the generally less important production attributes.
Taste and price were always included in the graded paired comparisons to provide a
common baseline against which the importance of the production attributes could be
evaluated.
Towards the end of the questionnaire, following some questions about demographic
and socio-economic characteristics, three choice sets with three partial pork profiles
each were presented to the respondents. Respondents were asked to choose one partial
pork profile from each choice set. The choices were used to assess the internal
predictive validity of estimated part-worths, or utilities (Hair et a!., 1998). These partial
profiles were based on production attributes only, i.e., they did not include taste and
price, and they were the same for all respondents.

Willingness to pay

The willingness to pay (WTP) part of the questionnaire started by establishing a respon-
dent's reference price for 'frequently consumed pork', thereby referring to answers
given in the introductory part of the questionnaire. Next, a respondent's WTP was
elicited for 'all consumer concerns' as well as for the four most important individual
attributes as derived from the CCA analysis. More specifically, we asked these WTPs
for pork "produced in such a way that the concerns, in the opinion of government and
consumer organizations, are dealt with according to the latest scientific developments".
WTPs were elicited in two different formats. From half of the respondents we
elicited the WTPs by asking how much they were willing to pay extra per kg for their
favourite kind of pork if the particular concerns would be met. From the other half we
elicited the WTPs by asking two questions about the total price per kg they were willing
to pay for their favourite kind of pork if the particular concerns would be met: (I) the
total price at which the respondent would certainly buy the pork, and (2) the total price
at which the respondent would no longer buy the pork. Respondents were randomly
assigned to one of the two formats. The first format ('extra') was chosen because in

NJAS 54-3, 2007 297

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

literature it is the standard format in which WTP-questions are asked. The second
format ('total certainly' and 'total no longer') aimed at better triggering respondents'
personal budget limitations.

Data collection

Data were gathered in the Netherlands in November 2001 over a period of three days.
At that time there were no major crises related to food safety or animal well being.
The most recent crisis, due to foot and mouth disease, was half a year before, in April
2001. In developing the questionnaire four focus group discussions were held with
people from different regions (rural and urban) in the Netherlands. Also, the question-
naire was pre-tested among 20 households. Then, the actual data gathering was done
with CentERpanel, a sample of 2000 households whose panel members frequently fill
in questionnaires through internet or set-top boxes (e.g. Donkers et a!', 2001). There
were 1444 respondents of which II99 fully completed the CCA analysis. Only this
group is considered in this paper. Although there were vegetarians and people who do
not eat pork, the sample of II9 9 only consisted of people consuming pork. Seventy-
five per cent of them were male, their average age was 47.3 years and 13% were older
than 65. Compared with the Dutch population, our respondents had on average more
children and a much higher income and education. Furthermore, the people in our
sample bought more expensive pork chops, bought more frequently at the butcher and
consumed (or stated to consume) relatively more labelled pork.

Segmentation analysis

The segmentation of our sample is based on the data from the CCA analysis. We used
a multivariate normal finite-mixture regression model (Wedel & Kamakura, 2000)
that was specially developed to simultaneously cope with the self-explicated data and
the graded paired comparisons (for similar models see Marshall & Bradlow, 2002 and
Ter Hofstede et a!', 2002). The model yields segment-level estimates of (I) part-worths
of the attribute levels, (2) a scaling factor relating the self-explicated ratings to the graded
paired comparisons, and (3) three variance parameters (one for the self-explicated
desirability ratings, one for the self-explicated importance ratings, and one for the
graded paired comparisons). Multivariate normal finite-mixture regression models
were fitted with 2 to 10 segments. Models were fitted, using the EM-algorithm, with
an alternating least-squares sequence in the maximization step (De Soete & Heiser,
1993; Wedel & Kamakura, 2000). To avoid local minima, the EM-algorithm was started
from 100 different random starts for each number of segments and the best solution
was retained.
In choosing the number of segments we used three criteria: a measure of internal
predictive validity, the Consistent Akaike Information Criterion (CAlC) as described by
Wedel & Kamakura (2000) and the interpretability of segments. The internal predictive
validity was assessed by predicting respondents' choices from the three choice sets
with the max-utility choice rule (Hair et a!', 1998), using the segment-level part-worth
estimates of the segment for which a respondent has the highest posterior probability.

298 NJAS 54-3, 2007

 Consumer preferences for pork attributes

For each number of segments, however, the predictive validity turned out to be between
40% and 45%, which is somewhat disappointing compared with the expected predictive
validity of 33.33% of a random choice model. These relatively low numbers might
be caused by the fact that choice sets were based on production attributes only (i.e.,
no taste and price). The second criterion, CAlC, suggests that a four-segment model
gives the best trade-off between the loglikelihood of the model and the number of
estimated parameters. However, there was a lack of interpretability of these segments
as they could not uniquely be classified. When comparing the interpretability of the
three to six-segment solutions it became clear that the six-segment solution revealed
interesting additional insights into the composition of the two largest segments of the
four-segment solution. Therefore, the results of the six-segment solution are reported
below.

Results
Preferences for pork chain attributes

The part-worth estimates from the CCA analysis for all attribute levels are shown in
Figure I for each of the six segments. For each attribute the highest part-worth indicates
the most preferred level and the lowest part-worth indicates the least preferred level.
For instance, for bone meal in pig feed not using it is preferred to using it. The further

Table 1. Perception scores per consumer segment of specific requirements for pork production in the
Netherlands, expressed as percentage of respondents agreeing with the statement I.

Statement Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
-------------------------- (%) ----------------------------

Pigs should only be kept 16.1 24.4* 2 13·4 24·3* 18·3 15·4
in designated areas
Production requirement 56 .1* 44. 8 52.9 4°·4 23. 2* 37. 0
should be above EU level
There should be no export 82·4* 66.2 75·4 67·4 50 .7* 58 ,9
oflive pigs
The number of pig farms 65·4* 49. 8 63. 6* 47. 2 45. 8 4 2.7*
should be reduced

I Percentages are for 'fully agree' or 'slightly agree' (other categories were 'slightly disagree', 'disagree',
or 'do not know').
2 An asterisk indicates a statistically significant (P <; 0.05) chi-square in cross-tabulation analysis and the
segments that contribute to this (I standardized residual 1" 1.96).

NJAS 54-3, 2007 299

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

Table 2. Socio-economic characteristics per consumer segment, expressed as percentage of

respondents, unless stated otherwise.

Characteristic Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
----------------------------- (%) ---------------------------

Males 77·5* 54. 1

Age (years)
< 35 18.5 ab 2 13.9 acdef 17.9 dghi 27.5 fi 28.0 beh
35- 6 5 71.7 62·7 64·7 56 .3 62.2
> 65 9. 8 23·4 17.4 16.2

Annual gross income (€)

< IO,OOO IO 14 ab 13 c 12 7b 9 ac
10,000-22,000 13 IO 13 14 8 8
22,000-33,000 21 26 24 22 26 21
33,000-44,000 16 22 19 20 22 22
> 44,000 40 28 32 32 37 40

Level of education
Low 6.3 abcde 25.0 afgh II.8 bf 18.3 c 12.9 eh 11.4 dgi
Medium 25·4 38 .5 34·4 36 . 2 30 .7 30 . 6
High 68·3 36 .5 53. 8 45·4 56 .4 58 .0

Profession
Industry 13.2 12·5 14.0 14. 2 23. 6 * 17.1
Government, services 4 6 .8* 27·5* 44. 6 * 34·4 30 .0 * 4 2 .9
Housewife, houseman 15.1 19·5* 12·9 15. 6 7,9* 14·3
Retired 12·7 26.0* 12·9 20.6 20·7 12.2*

---------------------------

No. of children per 0.80 0.56 ab 0.68 0.76 0.92 b 0.95 a

household 3

I An asterisk indicates a statistically different (P <; 0.05) chi-square value in cross-tabulation analysis and
the segments that contribute to this (I standardized residuall "1.96).
2 Percentages or numbers in the same row, followed by the same letter are statistically different
(P <; 0.05)·
The differences between percentages refer to differences between series of ordinal values (Mann-
Whitney U; a ~ 0.025).
3 Statistical differences are between arithmetic averages.

300 NJAS 54-3, 2007

 Consumer preferences for pork attributes

apart these part-worth estimates, the more important the attribute. For instance, when
comparing the distance between the most and least preferred levels of the attributes
bone meal in pig feed and residuals ofthe human food industry in pig feed the first one is
generally perceived as more important. Other issues on which consumers fairly agree
can be found at farm level: consumers generally prefer more space for pigs to less
space and housing in groups rather than individual housing. At processing and retail
level consumers agree on the relatively low importance of choice in quality and price
and they all prefer pork from the Netherlands, and a zero chance of Salmonella and
residues of medicines.
Besides these similarities, Figure 1 also shows differences between segments.
These differences formed the basis for classifying the segments. For instance, segment
1 (n ~ 205, 17%) is classified as ecologists. This segment is most strongly against
many non-ecological aspects of production, such as genetic modification in breeding,
herbicide residues in pig feed, irradiation for safety and medicine residues at retail
level. Also, at farm level they most strongly prefer pigs to have more space and they
have little against extra environmental requirements. In segment 2 (n ~ 201, 17%),
classified as tradition-minded, consumers typically prefer pigs to be raised in the
Netherlands and they like pig breeders to focus on increasing the variety among pigs.
At farm level, they prefer pigs to live on straw and they like them to be kept fully outdoors.
Segment 3 (n ~ 187, 16%) was classified as animal friends. They very much like pig
breeders to focus on disease resistance, and at farm level they attach a relatively high
importance to aspects of animal housing and handling. At the same time, at the
processing and retail level they find issues of taste and safety relatively less important.
Segment 4 (n ~ 218, 18%), was classified as the health-concerned, with health referring
to human health (zero Salmonella, no bone meal in pig feed) and pig health, and not
to environmental health. The latter is illustrated by the complete irrelevance of envi-
ronmental requirements at farm level. Segment 5 (n ~ 142, 12%) was called economists.
At the feed and breed level the economists attach little importance to issues of genetic
modification, and at farm level they favour tail docking, teeth clipping and castration.
At the processing and retail level they strongly prefer an excellent taste and they are
the only ones who favour a low price. Segment 6 (n ~ 246, 20%) was classified as
unpronounced, considering that the consumers had no specific requirements for any of
the chain attributes, not at the feed and breed level, nor at the farm or processing and
retail levels.
Answers to a number of 'pork-related citizen issues', presented at the end of the
questionnaire are mostly in line with the CCA outcomes and therefore support our
classification of the six segments (Table 1). For instance, ecologists are in favour of
enforcing production requirements above the ED ones, they are firmly against export
oflive pigs and they would favour a reduction in the number of pig farms. On the
other hand, the segments unpronounced and economists less often agree with these
Issues.
Not in line with CCA outcomes is the relatively high preference of the tradition-
minded segment for designated pig areas (24.4% agrees with the statement). This
segment, however, may have had a (wrong) idea of these areas. They may have thought
in terms of some large outdoor free-range areas for pigs, whereas after two major

NJAS 54-3, 2007 301

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

Table 3. Consumption characteristics per consumer segment.

Characteristic Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
--------------------------- (%) ----------------------------
Frequency ofpork
consumption (% resp.)
4-7 times per week 13.2 abc I 19.4 ad 16.6 21.6 b 30.3 d 19.9 c
I -3 times per week 54. 6 57. 2 58 .8 60.6 53·5 59·3
1-3 times per month 21.5 16,9 18.2 12.8 13·4 15.0
< once per month 8·3 4·5 5·3 4. 1 1.4 2.8
do not know 2·4 2.0 1.1 0·9 1.4 2.8
Pork chops most
frequently consumed (% resp.)
Shoulder (€ 6 per kg) 2 41.0 57. 2 47. 6 55. 0 53·5 45·9
Steak (€ 8 per kg) 4 0 .5 4 6 .3 28'9* 3 39·4 47·9 44·7
Cutlet (€ IO per kg) 28·3 35. 8 29·9 26.6 39·4 37·4
Filet (€ II per kg) 44·4 47·3 4 8 .1 4 6 .8 47·9 49. 2
Tenderloin (€ 12 per kg) 52 .7 60.2 57. 8 64. 2 66,9 63. 0
Other 20.8 15·4 14·4 16.1 11.4 21.8
Most frequent purchasing
place (% resp.) 4
Supermarket 66·3 64·7 69. 0 66.2 *
69·3 79·7
Butcher 27.8 32 .3 * 23·5 25·7 29. 6 18·3 *
-------------------------

Pork consumption per

label (% pork) 5
Regular pork (no label) 61.7 48.2 a 63·3 54·4 68.0 a 60.6
Integrated Chain Control 13.2 19.0 13·9 18,9 16·4 21.0

Free range 3. 2 6.6 2.1 3. 0 1.6 1.8

BestMeat 6 0·7 2.0 0·7 1.1 1.7 0·3
EKO (organic) 18.0 a 17·3 15·3 12·7 7.0 a IO.O

Environmental quality 2.2 3. 8 3·7 9. 1 4. 6 6·7

Biodynamic (organic) 1.0 a 3.1 abcde 1.0 b 0.7 c 0.8 e 0.3 d

I Percentages in the same row, followed by the same letter are statistically different (P., 0.05). The
differences between % respondents (% resp.) refer to differences between series of ordinal values
(Mann-Whitney U; a ~ 0.05).
2 The prices were not made known in the survey.
3 An asterisk indicates a statistically significant (P., 0.05) chi-square value in cross-tabulation analysis
and the segments that contribute to this (I standardized residual 1., 1.96).
4 Other suggested purchasing places, such as organic butcher, open market and farm, all scored very low.
5 Statistical differences are between arithmetic averages.
6 Non-existing pork label.

302 NJAS 54-3, 2007

 Consumer preferences for pork attributes

livestock epidemics these areas were planned to be confined industrial-like areas

comprising feed companies, large pig farms and slaughterhouses in order to reduce
the amount of transport throughout the Netherlands.
With respect to the socio-economic characteristics of the segments (Table 2),
ecologists were mostly female whereas economists were mostly male. With respect to
age, tradition-minded were somewhat older whereas unpronounced were relatively
young. The number of children and the level of income were highest for the segments
unpronounced and economists. With regard to the level of education, ecologists turned
out to be relatively highly educated. With respect to people's profession, we found that
industry relatively often employs economists, and governments employ ecologists and
animal friends. Tradition-minded consumers were relatively often found in the categories
housewife/houseman and retired. Hardly anybody worked in the agricultural sector or
had any close relationship with it (not shown in Table 2).
In order to further characterize the six segments, people's stated consumption of
fresh pork was analysed (Table 3). Ecologists consumed pork less frequently than
tradition-minded, health-concerned and unpronounced. With respect to the consumption
ofvarious types of pork chops, segments hardly differed. With regard to pork consumption
per label, all segments mostly consumed non-labelled pork, ranging from 48.2% for
tradition-minded to 68.0% for economists. EKO pork was consumed most by ecologists:
18.0% of their total pork consumption. Biodynamic pork was consumed most by
tradition-minded (3.1%). The fake label BestMeat was hardly stated to be consumed at
all, which indicates proper awareness among respondents.

Knowledge about pork production

People's perceptions about pork production or specific pork labels might be influenced
by their knowledge about aspects such as pigs' living circumstances and pork production
techniques. We therefore recorded consumers' knowledge about such issues just
before the CCA analysis. Four multiple choice questions and eight statements were
used. These statements had to be evaluated twice: once for regular pork and once for
EKO pork. Table 4 shows the percentages of correct answers that were lowest for the
statements ('do not know' was considered an incorrect answer).
In relation to the multiple choice questions, people appeared to know more about
the age of slaughter pigs and the percentage of pigs kept outside than about farm-gate
price levels and the number of litters per sow per year. (More than one third of the
respondents believed sows to give birth four times a year.) With respect to the statements
there were, for both regular and EKO pork, only three issues to which more than half
of the consumers gave the correct answer. For regular pork these were the aspects
about not having the obligation to use organic feed, to enable pigs to go outside and to
only use homeopathic medicines. For EKO pork more than half of the consumers only
knew about feeding organic feed, using feed without antibiotics and the obligation to
provide pigs with outdoor facilities. However, most consumers did not know that with
EKO pork the use of homeopathic medicines is not compulsory either. Four other
items about which consumers hardly knew whether regular pork and EKO pork were
comparable included (I) that not only ingredients from their own country can be used

NJAS 54-3, 2007 303

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

in the feed, (2) that the use of hormones is not allowed, (3) that irradiation of meat is
not allowed, and (4) that artificial insemination is allowed.
Knowledge scores (lower part of Table 4) are based on all statements; the Cronbach's
alpha (Hair et a!., 1998) for the statements on regular and EKO pig farming systems
was 0.82 and 0.90, respectively. Based on the knowledge scores, we conclude that
tradition-minded had relatively little knowledge about regular pig farming and that none
of the segments was discriminating on knowledge with respect to EKO farming. Also,
pair-wise comparisons (not indicated in Table 4) show that all segments, except tradition-
minded and health-concerned, knew more about regular pig farming than about EKO
farming.

Table 4. Knowledge per consumer segment about pork production in the Netherlands, expressed as
percentage respondents (% resp.) giving the correct answer.

Items Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
--------------------------- (%) ---------------------------

Correct answers to
multiple choice questions 1

(% resp.)
1. Age of slaughter pig 95* 2 86* 9° 9° 9 6* 84*
(6 months)
2. Pork price at farm gate 67 58 61 61 76* 63
(€ 1.36 per kg)
* 86* 88* *
3. Percentage pigs raised 95 91 94 95
outdoors (5%)
4. No. oflitters per sow 56 60 55 58 60 54
per year (2)

Correct answers to
statements 3 (% resp.)
1. Only feed without 25 - 72* 22 - 56 * *
34 - 67 20 *- 61 23 - 62 26 -70
antibiotics (no 4 - yes)
2. Only ingredients from 45 - 28 29 * - I 8* 45 - 22 4 1 - 25 44 - 26 4 2 - 31
*
own country (no - no)
3. Compulsory use of 80 *- 75 * *
59 - 63 66 - 73 62 - 62 * 68 - 63 78* - 72
organic feed (no - yes)
4. Pigs must be able to 83 *- 62 63 * - 57 69 - 64 64 * - 55 74 - 52 80 * - 58
go outside (no - yes)
* 12 * - 26
5. Artificial insemination 4 - 18 II - 22 7 - 24 9 - 20 9 - 24
allowed (yes - yes)

304 NJAS 54-3, 2007

 Consumer preferences for pork attributes

Table 4 (cont'd)

Items Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
--------------------------- (%) ---------------------------
Correct answers to
statements 3 (% resp.)
6.0nly use of homeopathic 76* - 32 60 * - 22 69 - 28 62 * - 26 69 - 24
medicines (no - no)
7. Use of hormones 32 - 6* 29 - I 6* 36 - 12 29 -9 25 - 8
allowed (no - no)
8. Irradiation of meat 27 -8 25 - 15 35 - 12 3° - '4 28 - 8
allowed (no - no)

Knowledge scores 5
Regular pork 3.7 a 6 3.0 abc
EKO pork 3-43. 6 3·73·7

1 The correct answers are given in brackets. The possible answers for I: 6 weeks, 6 months, 18 months;
for 2: € 1.36 per kg, € 2.73 per kg, € 5.45 per kg; for 3: 5%, 50%, 95%; for 4: 2,4, 6.
2 An asterisk indicates a statistically significant chi-square value in cross-tabulation analysis and the
segments that contribute to it (I standardized residual 1" 1.96).
3 The correct answers (in brackets) are for regular and EKO pork, respectively. The possible answers
were: yes, no, do not know.
4 Although debates about abandoning the use of antimicrobial growth promoters in feed already started
in the '99os, the actual prohibition was only enforced as of I January 2006 (EU, 1813/2003, article II).
5 Totalized scores for statements based on a value I for each correct answer and ° otherwise, including
'do not know'.
6 Scores followed by the same letter are statistically different (P <; 0.05).

Willingness to pay for improvements in pork production

Table 5 summarizes the results of our WTP analysis. Consumers' reference prices
for pork were between € 7.90 per kg for tradition-minded and € 9.00 per kg for ecol-
ogists. On top of the reference prices and given that 'all concerns' would be properly
addressed, people were willing to pay from € 2.90 (unpronounced) to € 5.20 per kg pork
(tradition-minded). In case individual concerns as derived from the CCA analysis would
be met, these amounts ranged from € 4.00 per kg for unpronounced to € 7.00 per kg
for animal friends. For all segments the extra price on top of the reference price was
highest for dealing with individual concerns. Only for economists the individuality of
the concerns did not influence their willingness to pay. With respect to the total prices,
'certainly' prices were below the 'no longer' prices. However, some of the certainly

NJAS 54-3, 2007 305

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

Table 5. Pork prices per consumer segment and willingness to pay (WTPj, addressing 'all concerns' or
'individual concerns'.

Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
------------------------- (€ per kg) -----------------------

Reference price 9. 0 7·9 8·4 8.1 8.1 8·5

WTP for all concerns

*I 5. 2 * a 2
Extra price on top of 3·7 4.8 * 5.1 * b 3·4 2·9 * ab
reference price
Total price 'certainly' 8.0 6,9 ab 8.8 * ac 8.0 6,9 cd 8.5 * bd
Total price 'no longer' 9. 6 8.4 a ro.I *b 8.7 c 7.8 bd ro.4 acd

WTP for individual concerns

Extra price on top of * 6·5 * a 7.0 * bc 6.3 * d 4.2 c 4. 0 * abd
4·9
reference price
Total price 'certainly' 8·7 7.1 ab 9.8 * a 8·3 7·9 9.1 *b
Total price 'no longer' 9. 8 8.5 a 11.3 * a 9·3 8·5 11.0

1 Prices in the same column, followed by an asterisk are statistically different (P <; 0.05) for similar WTP
framings.
2 Prices in the same row, followed by the same letter are statistically different (P <; 0.05).

prices were also below reference price levels. Some people clearly did not exactly recall
their own reference price - or believed that the price they usually paid was too high
and that they now had a chance to express that. Total prices were relatively high for
animal friends and unpronounced.
Instead oflooking at average willingness-to-pay numbers as presented in Table 5,
it is also interesting to consider the underlying range of values. This has been worked
out for WTPs addressing individual concerns and only for extra and certainly numbers.
In order to directly compare these WTPs, both have been recalculated to percentages on
top of an individual's reference price (Figure 2). We distinguished between four WTP-
categories starting at 0% (including WTPs below 0% in case consumers indicated a
total price below their own reference price) up to WTPs above 75% of an individual's
reference price.
The upper part of Figure 2 illustrates that in case of the 'extra framing' more
consumers were found in the 25-75% and> 75% WTP-categories than in the case of
the 'certainly framing'. With extra framing these categories in all segments comprised
more than half of the consumers, whereas with certainly framing this was reduced to
less than half of the consumers, again in all segments.

306 NJAS 54-3, 2007

 Consumer preferences for pork attributes

50 WTP-Extra
45
40
!l 35
..
c:
Ql
"'C
c: 30
0
....
t
~25
~
"ji. 20
15
..
::::
.. ~~
::::
::::
10
5
..
I.
:.:
;:::
::::
:.:.
:~:~
0
50
WTP-Celtainly -:::
45
40

~ 35
Ql
~:
"'C
c: 30
8. 25
'"~ 20 .. 7'
"ji. ..
15
10 .. ..
5
0
R:J .. m .. m :i __ ,-~
Ecologist Tradition-mlndedAnimal friend Health-eoncemed Economist Unpronounced

~ <=0% o >0-25% ~>25-75% t8 >75%

Figure 2. Percentage respondents per segment in the willingness-to-pay (WTP) categories 'extra' and
'certainly' for pork prices on top of reference prices. WTPs reflect individual concerns.

Table 6. Consumer segment's opinion about the financial contribution of consumers, government and
polluters to improving farm-level pork production methods, expressed as percentage of respondents.

Category Consumer segment

Ecologist Tradition- Animal Health- Economist Unpronounced

minded friend concerned
----------------------------- (%) ---------------------------

Consumers 34.4 A I 29.9 a 2 35.6 A 31.2 b 40.8 abcAB 32.9 c

Dutch government 24.6 abAB 30.9 aA 26.6 AB 31.6 bc 24.6 cA 29.3 A
Polluters 3 39.0 aB 34.8 A 35.9 bB 36.1 c 28,9 abcdB 35.9 dA

I Prices in the same column, followed by a common capital letter are statistically different (P <; 0.05).
2 Prices in the same row, followed by a common lower cast letter are statistically different (P <; 0.05).
3 Polluters are defined here as pig farmers who do not improve their way of farming.

NJAS 54-3, 2007 307

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

0.7 EcoIogIIl TIOlitl.....-indld

0.6

1 0.5
jO.4

JO.3

_-
0.2
0.1
0.0
0.7
_lllIond .....
0.6

p.5
t 0.4
p-3
0. 0.2
0.1
0.0
0.7
E<onomIIl Unp<onoun<td
0.6

IDS
.

1
cD.4
0.3
0.0.2

0.1
0.0
FS AW EN SO NA EX PO OP FS AW EN sa NA EX PO OP

C 0 label a IKB D Free range EI BestMc8I • EKO [!) Environmcnllll quality EI Biodynamic

Figure 3. Perception scores per segment for different pork labels and for the consumer concerns food
safety (FS), animal welfare (AW), environment (EN), sensory quality (SQ), naturalness (NA), expensiven-
ess (EX), price/quality ratio (PQ) and overall preference (OP).

The considerable number of consumers who indicated a WTP of"" 0%, especially in
the certainly framing, might, besides reasons indicated above, include consumers
who are concerned about pork production issues but who perceive that someone else
should pay for improving these. Towards the end of the questionnaire we asked
respondents to indicate their preferred financial contribution from consumers, Dutch
government and 'polluters' for improving farm-level pork production methods. The results
(Table 6) show, although there is indeed a role for other parties such as government
and polluters, that segments also saw a role for consumers themselves.

Pork label perception

After having indicated consumption figures per pork label, as shown in the lower part
of Table 3, respondents were asked about their pork label perception. Labels had to be

308 NJAS 54-3, 2007

 Consumer preferences for pork attributes

evaluated with respect to animal welfare, food safety, environment, naturalness, sensory
quality, expensiveness, price/quality ratio and overall preference. Respondents were
asked to indicate the label perceived to score best, second best and third best for each
specific item. Answers were dichotomized for subsequent analyses, i.e., a label scores
1 on an item if the label is part of a respondent's top 3, and 0 otherwise (including 'do

not know'). Perception scores per segment and label are shown in Figure 3.
Regular pork (no label), which was stated to be consumed most (Table 3), scored
high on some items (price/quality ratio and overall preference) and for some segments
(especially economists) but EKO (organic) generally scored higher (Figure 3). This mostly
also held for all other items.

Discussion and conclusions

This paper aimed at identifying consumer segments for pork, based on multiple
attributes of the entire pork supply chain. Based on an extensive customized conjoint
analysis (CCA) and by developing a finite-mixture regression model, we identified six
segments of pork consumers classified as ecologists (17%), tradition-minded (17%),
animal friends (16%), health-concerned (18%), economists (1Z%) and unpronounced (zo%).
Subsequently, these segments were described on issues ranging from (standard)
socio-economic characteristics and pork consumption figures to 'pig farming knowledge
scores' and willingness-to-pay numbers for improving pork production practices.
Besides, segments were evaluated on a number of 'pork-related citizen questions',
such as whether the Netherlands should stop exporting live pigs and whether production
requirements in the Netherlands should be above EU-level.
The segment unpronounced was not really interested in any of the production items
studied (although they had a rather good knowledge of them), and had no strong overall
preference for any of the current labels. They mostly bought pork in the supermarket
and were willing to pay relatively small amounts of money for production improvements.
All other segments were found to be interested in at least some of the production
items. For instance, economists strongly preferred cheap but excellent tasting pork of
Dutch origin. Furthermore, they perceived farm-level traceability as relatively unimportant,
did not seem to care about genetic modification in feed or animal breeding and had
the highest overall preference for regular pork without any label.
Due to the large number and detailed character of the attributes, our results cannot be
directly compared with literature. However, generally, findings are in line with other
multi-attribute studies. For instance, Verbeke & Viaene (1999) also indicated the high
relevance for Belgian consumers of acceptable pork production methods. With respect
to the country of origin, also Dransfield et al. (zooS) concluded that pork consumers
in France, Denmark, Sweden and the UK mostly appreciate pork labelled as 'home
produced' or 'outdoor'. In relation to the importance of safety issues, Krystallis et al.
(zo06) identified 'health clusters' as well.
For decision-makers in pork supply chains to start developing specific pork chains
and labels based on the attributes and issues studied, we deem four questions crucial:
1. Are segments valid? Are consumers able to distinguish between pork production attributes?

NJAS 54-3, 2007 309

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

To begin with, we found that consumers clearly had different preferences for a number
of production attributes. Especially issues of animal welfare and food safety as well
as farm-related aspects in general appeared to trigger people's explicit preferences.
Also, answers to preference questions in the CCA analysis were mostly consistent with
answers to willingness-to-pay and label perception questions. We therefore conclude
that the identified segments are rather robust and that consumers are indeed able to
express their preference for detailed issues of pork production. On the other hand, we
did not have a representative sample, as it was biased towards more wealthy people
with higher levels of education and income and with more children than average. Also,
there were relatively many male respondents, whereas the daily shopping is mostly
done by women. Both issues, however, are not likely to change the existence of the
segments, although they might change somewhat their size. Another concern could
be that respondents had likely been giving some socially desirable answers, at least in
some parts of the questionnaire, as they stated to consume a rather high percentage
oflabelled pork (40% to 50%), whereas in reality market shares for labelled pork were
less than 1%. However, segments are based on data from the CCA analysis, which is
specifically designed to reduce problems of overstatements and socially desirable answers.
2. Can segments be sufficiently targeted? Given the specific preferences and the socio-

economic and consumption characteristics per segment we believe that adequate

targeting must be possible. However, with relatively low knowledge scores and some
absurd ideas about current pork production and safety issues, care should be taken
about how to approach and communicate with consumers. Attention should for
instance be paid to the (widespread) import of feed ingredients, the (image of) exporting
live pigs, the prohibition to use hormones and irradiation, and the relatively low prices
for pigs at farm level. It should also be noted, however, that it is not only 'facts and
figures' knowledge that can bring a good image to a label, as illustrated by our EKO
pork findings.
3. What is the financial room for investments? A pessimistic decision-maker would point
to the high preference for current reference prices, the high percentage of people in
the"" 0% willingness-to-pay categories and the relatively modest role of consumers in
financing improvements in pork production methods. But more optimistic decision-
makers can see that most segments hardly cared about somewhat higher prices per
kg of pork and that a considerable number of consumers were certainly willing to pay
more than 25% on top of their current reference price provided that their individual
concerns about pork production would be met. This should provide ample financial
room for investments, thereby noting that many investments might even be rather
subtle given the relatively high perception scores for regular and IKB (bulk) pork,
especially on the price/quality item.
4. What are the practical implications for chain design: six different chains or synergy?
Our finding that consumers' willingness-to-pay numbers were highest when individual
concerns were targeted is a strong argument in favour of six different chains.
However, there are also arguments in favour of attempting to achieve synergy as: (I)
consumers had difficulties with distinguishing between different labels, as was illustrated
by our label perception analysis, (2) current legislation reduces differentiation oppor-
tunities, i.e., growth promoters in feed, genetic modification in breeding, pork irradiation

310 NJAS 54-3, 2007

 Consumer preferences for pork attributes

techniques and 'no traceability' are already forbidden, and the level of genetically
modified substances in feed and residues of medicines in pork are already strictly
limited; and (3) differences mainly exist at farm level (living surface, outdoor housing,
pig handling).
Our main recommendations for decision-makers in pork supply chains can be
summarized as follows:
1. Pork should not be marketed as a bulk product. There is a basis for segmentation

in the market with distinct requirements for pork and the way it is produced.
2. Although there is financial room for investments into each of the six segments

identified, a differentiation up to six labels may be too much. From the consumer
perspective it is not likely that people can differentiate between six labels. From the
producer perspective the current legal framework does not even allow to respond to
each of the segments.
3. The market of fresh pork entails, besides necessary 'business-to-business transpa-
rency', a viable opportunity for 'business-to-consumer transparency', but care should
be taken in communicating with consumers.

Acknowledgements
The research described in this paper is part of the Green Piggery project of
Wageningen University and the Institut National de la Recherche Agronomique
(INRA), France.

References
Anonymous, 2002. Regulation (EC) No 178/2002. European Commission, Brussels.
<http://europa.eu.int/eur-lex/pri / en / oj / dati 2002 /L03 I /L03 12 002 02 01enooo I 0024.pdf>
Anonymous, 2006. Statistics Netherlands 2006. <http://www.cbs.nl> Accessed 3 August 2006.
Boehlje, M.D. & D.A. Lins, 1998. Risks and risk management in an industrialised agriculture.
Agricultural Finance Review 58: 1-16.
Den Ouden, M., J.T. Nijsing, A.A. Dijkhhuizen & R.B.M. Huirne, 1997. Economic optimization of pork
production-marketing chains. I. Model input on animal welfare and costs. Livestock Production
Science 48: 23-37.
De Soete, G. & W.J. Heiser, 1993. A latent class unfolding model for analyzing single stimulus
preference ratings. Psychometrika 58: 545-565.
Donkers, B., B. Melenberg & A. Van So est, 2001. Estimating risk attitudes using lotteries, a large
sample approach. Journal of Risk and Uncertainty 22: 165-195.
Dransfield, E., T.M. Ngapo, N.A. Nielsen, L. Bredahl, L., P.O. Sjoden, M. Magnusson, M.M. Campo &
G.R. Nute, 2005. Consumer choice and suggested price for pork as influenced by its appearance,
taste and information concerning country of origin and organic pig production. Meat Science
69: 61-70 .
Hair, J.F., R.E. Anderson, R.L. Tatham, R.L. & W.e. Black, 1998. Multivariate Data Analysis (5th
edition). Prentice Hall, New Jersey, 730 pp.

NJAS 54-3, 2007 311

M.P.M. Meuwissen, I.A. Van Der Lans and R.B.M. Huirne

Hensel-Borner, S. & H. Sattler, 2000. Validity of customized and adaptive hybrid conjoint analysis. In:
R. Decker & W. Gaul (Eds), Classification and Information Processing at the Turn of the
MILLENIUM. Proceedings of the 23rd Annual Conference of the Gesellschaft fUr Klassifikation,
IO-14 May 1999, Bielefeld. Springer, Berlin, pp. 320-329. <http://www.henriksattler.de/publikatio
nen/H S_04o.pdf>
Hoffmann, R., 2000. Country of origin - a consumer perception perspective of fresh meat. Food
Journal I02: 2II-229.
Huber, J. & K. Zwerina, 1996. The importance of utility balance in efficient choice designs. Journal of
Marketing Research 33: 307-317.
Krystallis, A & LS. Arvanitoyannis, 2006. Investigating the concept of meat quality from the
consumers' perspective: the case of Greece. Meat Science 72: 164-176.
Marshall, P. & E.T. Bradlow, 2002. A unified approach to conjoint analysis models. Journal of
American Statistics Association 97: 674-682.
Meuwissen, M.P.M. & LA Van Der Lans, 2005. Trade-offs between consumer concerns: an application
for pork supply chains. Acta Agriculturae Scandinavica, Section C - Food Economics 2: 27-34.
Meuwissen, M.P.M., AG.J. Velthuis, H. Hogeveen & R.B.M. Huirne, 2003. Traceability and certifi-
cation in meat supply chains. Journal ofAgribusiness 21: 167-181.
Nayga, R.M., 1996. Sociodemographic influences on consumer concern for food safety: the case of
irradiation, antibiotics, hormones, and pesticides. Review ofAgricultural Economics 18: 467-475.
Ngapo, T.M., E. Dransfield, J.-F. Martin, M. Magnusson, L. Bredahl & G.R. Nute, 2003. Consumer
perceptions: pork and pig production, insights from France, England, Sweden and Denmark. Meat
Science 66: 125-134.
Schifferstein, H.N.J., M.J.J.M. Candel & H.C.M. Van Trijp, 1998. A comprehensive approach to image
research: an illustration for fresh meat products in the Netherlands. Journal of Social Research in
Agriculture (TSL) 13: 163-175.
Srinivasan, V. & C.S. Park, 1997. Surprising robustness of the self-explicated approach to customer
preference structure measurement. Journal of Marketing Research 34: 286-291.
Ter Hofstede, F., K. Youngchan & M. Wedel, 2002. Bayesian prediction in hybrid conjoint analysis.
Journal of Marketing Research 39: 253-261.
Verbeke, W., 2001. Beliefs, attitude and behaviour towards fresh meat revisited after the Belgian dioxin
crisis. Food Quality Preference 12: 489-498.
Verbeke, W. & J. Viaene, 1999. Beliefs, attitude and behaviour towards fresh meat consumption in
Belgium: empirical evidence from a consumer survey. Food Quality Preference IO: 437-445.
Wedel, M. & W.A Kamakura, 2000. Market Segmentation: Conceptual and Methodological
Foundations. Kluwer, Boston, 382 pp.
Williamson, O.E., 1985. The Economic Institution of Capitalism: Firms, Markets, Relational
Contracting. The Free Press, New York, 450 pp.
Yussefi, M., 2006. Organic farming worldwide 2006: overview and main statistics. In: H. Willer &
M. Yussefi (Eds), The World of Organic Agriculture: Statistics and Emerging Trends. International
Federation of Organic Agriculture Movements (IFOAM), Bonn, pp. 23-37.
<http://orgprints.org/ 51 6 I / ° /yussefi- 2006-overview.pdf>
I

312 NJAS 54-3, 2007