Food Policy 37 (2012) 232–245

Contents lists available at SciVerse ScienceDirect

Food Policy
journal homepage: www.elsevier.com/locate/foodpol

Food choice, health information and functional ingredients: An experimental

auction employing bread
Nicole Elizabeth Hellyer a, Iain Fraser b,d,⇑, Janet Haddock-Fraser c
a
Kent Business School, University of Kent, Canterbury CT2 7PE, United Kingdom
b
School of Economics, University of Kent, Canterbury CT2 7NP, United Kingdom
c
Faculty of Social and Applied Sciences, Canterbury Christ Church University, Canterbury CT1 1QU, United Kingdom
d
School of Economics and Finance, La Trobe University, Melbourne, Victoria 3083, Australia

a r t i c l e i n f o a b s t r a c t

Article history: In this paper we present the results of an experimental auction conducted to examine the influence of
Received 14 October 2010 health and nutritional information on food choice and in particular estimate consumer willingness to
Received in revised form 23 September pay (WTP) for bread that contains functional ingredients. Employing a sandwich we find that consumers
2011
are WTP more for a whole grain and whole grain granary bread sandwich than other bread types includ-
Accepted 13 February 2012
Available online 20 March 2012
ing white bread that contains a functional ingredient. We also find that consumers react positively to the
provision of nutritional and health benefit information but that this effect occurs regardless of whether
we supply specific or non-specific health benefit information. We discuss information provision and
Keywords:
Functional foods
health policy implications that emerge from our analysis for bread products in the sandwich market.
Fibre Ó 2012 Elsevier Ltd. All rights reserved.
Bread
Whole grains
Willingness-to-pay

Introduction Haddock-Fraser, 2011). It has been estimated that a total of

There are an increasing number of novel food products being
developed and offered to the market, which are frequently differ-
entiated by the modified attributes they offer the consumer. Many
of these are being marketed in terms of the benefits they offer for
consumer health as well as the potential to reduce the risk of dis-
eases. Some of these new food product developments have been la-
belled as ‘functional foods’ (e.g. Siró et al., 2008).
In principle, functional foods have the potential to improve pop-
ulation health in line with the objectives identified by national
public health strategies, because the increased consumption of
functional ingredients in bakery products implies that the quantity
of fibre in the diet, especially whole grains, increases. Whole grain
cereals are an important component of a healthy balanced diet
containing a range of macronutrients (fat, carbohydrates and fibre)
and micronutrients (vitamins, minerals and phytonutrients)
(Dewettinck et al., 2008). Consuming whole grain foods has been
shown to prevent several different, fatal and prominent non-com-
municable diseases, such as Type 2 Diabetes Mellitus, Cardiovascu-
lar Disease and certain cancers (Anderson, 2003, and Hellyer and
⇑ Corresponding author at: School of Economics, University of Kent, Kent,
Canterbury CT2 7NP, United Kingdom. Tel.: +44 1227 823513.
E-mail addresses: i.m.fraser@kent.ac.uk (I. Fraser), janet.haddock-fraser@canter
bury.ac.uk (J. Haddock-Fraser).
113,836 deaths per year in the UK could potentially be reduced
by the consumption of whole grains and the alteration of lifestyle
choices (Mozaffarian et al., 2008). However, consumers are reluc-
tant to alter their eating habits, even when they know that there
maybe detrimental consequences to their health in the future
(Williamson et al., 2000).
Although consumers and health professionals have displayed
positive preferences for functional food in general, there is evi-
dence that consumers differ in the extent to which they buy spe-
cific food products with functional ingredients, especially bakery
products. Bread has recently become a vehicle for functional ingre-
dients, for example, with the introduction of omega 3 fatty acids. In
addition, whole grains naturally contain functional ingredients,
including phytochemicals (phytic acid, glutathione, and phytoster-
ols), as well as dietary fibre (Inulin and beta-glucan), which yield
health benefits, which could not be consumed in white bread alter-
natives (Sidhu et al., 2007).
Although there is little difficulty including functional ingredi-
ents in bakery products, whether the resulting functional product
meets consumer demands is less clear. Compared to dairy products
current consumption of bakery functional products, specifically
bread, is relatively low. Indeed, even with advances in bread as a
‘functional food’ there are various reasons why consumers might
be unwilling to adopt these products. Siró et al. (2008) observe that
the acceptance of functional foods is conditional on the product

0306-9192/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodpol.2012.02.005
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
and data indicates that consumers are less likely to eat whole grain
bread compared to other types of grain (Hoare et al., 2004). To
combat the issues of non-adoption one approach the food manu-
facturing industry can take is to include beneficial functional ingre-
dients (e.g., Inulin) within white bread.
This research examines the effect of different levels of informa-
tion through the acceptance and willingness to pay (WTP) using an
experimental food auction. The auction required participants to bid
over a selection of sandwiches made from different types of bread.
The experiment was designed to focus on differences in (i) respon-
dents’ WTP for different types of bread; (ii) the level of information
relating to health impacts bids; and (iii) the specific health claim ef-
fects the bids received. Even though experimental auctions are lab-
oratory based, real products and money are exchanged, providing
the individual an incentive to reveal the real value of the product
being studied and as a result reduce excessive hypothetical bias.
In addition, having consumers pay money during the auction the
bids placed may better reflect actual preferences than an attitudinal
survey, as there are consequences to over and underbidding.
There already exist several papers in the literature that have
developed conceptual models to explain the relationship between
consumer valuation of products and information provision in rela-
tion to food (e.g., Ippolito and Mathios, 1990; Teisl et al., 2001, and
Lusk et al., 2004). Ippolito and Mathios provided an important
seminal contribution to this literature that subsequent papers
draw upon. For example, the model presented by Lusk et al. exam-
ines how consumers’ valuation changes as a result of receiving po-
sitive information. However, like Teisl et al. we note that the value
of additional information should be greater for an individual who
has a lower level of knowledge prior to the new information being
revealed.
The structure of the paper is as follows. In Section 2 we briefly
review the literature on health and food with a specific focus on
bread as well as functional ingredients. In Section 3 we describe
and explain our experimental auction. Then in Section 4 we exam-
ine the data generated by our auction and present the results of our
analysis. Finally, in Section 5 we conclude.
Literature review
There is a rapidly growing literature on the importance of func-
tional ingredients (e.g., Cox et al., 2007; Siró et al., 2008), and in
particular fibre and whole grains in relation to consumer health
(e.g., Bitzios et al., 2011). There is also an unrelated literature on
the design, implementation and evaluation of experimental auc-
tions in relation to food and health (Lusk et al., 2004; Rozan
et al., 2004).
Before we consider the antecedent literature on experimental
auctions it is important to understand how health and nutritional
information can influence consumer choice in relation to func-
tional ingredients, whole grains and bread. As would be expected
there are many diverse papers dealing with this topic in the liter-
ature. A number of these studies involve the consumption of food
products as part of the experimental design. For example, Mialon
et al. (2002) assessed responses to the provision of information
about the fibre content of bread and English muffins. Participants
were asked to rate six products on acceptance, sensory, health
and nutrition variables. They found that a healthy perception about
nutrition could be influenced by written information. Kihlberg
et al. (2005) examined four different bread products based on the
production of the flour. They found that a health claim about cho-
lesterol only became significant to the reported liking of the bread
product when it was combined with the flour’s origin. Also liking
increased for bread produced from conventional flour but not for
the organic flour. Annett et al. (2008) also examined consumer
233
acceptance of different flour types (i.e. organic and conventional)
and the effect of health information. Their results demonstrated
that health information in combination with sensory evaluation in-
creased liking for organic bread, whereas environmental informa-
tion had no effect. In these papers participants have not had to
purchase the products being examined. Also many whole grain
food products already have health claims and so comparison of
no health information with health information is no longer practi-
cal and so the differences in the claims used should be examined.
Mancino et al. (2008) provide a different set of insights by
examining why consumers increased their consumption of whole
grains in the US after the publication of the 2005 Dietary Guide-
lines. The insight they provide draws on Ippolito and Mathios
(1990), in that it is suggested that public policy lead to food man-
ufacturers introducing new and differentiated products that con-
sumers readily adopted. However, Golan and Unnevehr (2008)
note that competition between food manufacturers in terms of
health attributes need not result in healthier food products enter-
ing the market place.
We now turn to the experimental auction literature where there
have been many food related applications (e.g. Poole et al., 2007,
and Rousu et al., 2007). An advantage of employing an experimental
auction is that the provision of information to participants can be
isolated from other effects. As a result, auctions have proven popu-
lar as a method to examine how information provision impacts the
potential WTP of consumers for food products. For example,
Marette et al. (2010) consider how information about functional
ingredients can influence WTP for a fortified yoghurt drink. Their
results show a positive impact of information that details choles-
terol reducing properties. Importantly, they find this effect for
participants regardless of having cholesterol problems themselves.
Also they note that auction participants place less emphasis on the
negative impacts and more on the positive impacts of a functional
food. This observation relates to the findings of Naylor et al.
(2009) and confirmatory bias.
Turning to a bread-specific example, an interesting auction
experiment that employed baguettes is presented by Rozan et al.
(2004), who examined consumer responses to information relating
to food safety. Bread was selected for use in this auction because it
is among the most frequently purchased items by households.
Employing the second-price Vickrey auction and the Becker–deG-
root–Marschak (BDM) procedure they found that product prices
increased if the product is certified as safe, in this case from expo-
sure to heavy metals. Interestingly, Rozan et al. found that the BDM
procedure yielded higher bids than the second-price Vickrey auc-
tion, which is in contrast to the earlier findings reported in the lit-
erature. Ginon et al. (2009) also employ baguettes, except they
examine how nutritional information impacted WTP. The experi-
ment provided specific information about fibre content and then
additional information about the nutritional effects of fibre. Like
Rozan et al. they employed the BDM. The main finding of this re-
search was that nutritional information about potential health
benefits did not induce a significant change in WTP. But, when
the bread was explicitly labelled ‘‘source of fibre’’ this did increase
WTP, although the magnitude of the effect was small.
Another informative study by Hobbs et al. (2006) estimated
consumer WTP for bison compared to beef, with bison offering
health-related attributes including being lower in fat. Employing
sandwiches as the food to deliver the two types of meat, employing
a Vickrey second-price auction, they found no significant differ-
ence in WTP for bison (a luxury item) over beef. Indeed, ensuring
consumers had a positive eating experience was in fact more
important than the health information provided. Another auction
employing sandwiches was conducted by Drichoutis et al. (2008).
This study used three types of sandwich: 6-in. sub sandwich; a
wrapped pita sandwich; and a Mediterranean type sandwich.
234 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
Employing a Vickrey second-price auction, in round one, partici-
pants could examine and taste the sandwiches. In the next round
they undertook the same activity except this time nutritional infor-
mation was provided. They estimate WTP by employing a Tobit
specification with upper and lower limits, finding that the provi-
sion of information matters and is indicated in their results via
higher WTP estimates.
Gaps in the literature
Research within the literature has found that in general individ-
uals are WTP for functional food or food containing functional
ingredients as long as other important properties of the product
are not compromised e.g. hedonic characteristics. In addition,
experimental auctions have shown that the effect of information
provision relating to health benefits associated with a food product
can yield an increase in WTP as long as other important properties
are not compromised.
It is also important to note that the type of information treat-
ment employed in experimental auctions is frequently unrealistic,
as the auction moves from a situation of no written information to
full information. Additionally existing food products will be subject
to an evolution of information, such as, basic nutritional informa-
tion, specific nutritional information and then additional health re-
lated claims.
Another limitation of existing experimental auctions, especially
those that consider food products is that they have not attempted
to control for issues related to eating behaviours. It would also ap-
pear to be the case that no effort has been made to control for sub-
jects level of hunger or mood whilst participating in the auction.
Both of these limitations are addressed in the experimental auction
reported in this paper.
Experimental auction design and implementation
Auction design
Three focus groups attended by 14 people in total informed the
development of the food auction process, including the order of the
auction rounds (seeing the product, nutritional information and
tasting to be similar to a purchase situation) and the selected sand-
wich filling. They were recruited from staff and students at the
University of Kent and advertised using various mailing lists. An
incentive of tea, coffee and biscuits was provided to all those that
attended.
As a result of the focus groups a cheese ploughman’s sandwich
was identified as the most suitable for the food auction, because it
did not prohibit the participation of vegetarians. Other sandwich
fillings were discussed including individuals having a choice of
their preferred sandwich. However, this was not a possibility due
to logistical restrictions of having to prepare the sandwiches on
the morning of the auction. The choice of the cheese ploughman’s
sandwich is also supported by information presented by the British
Sandwich Association (2009) who identified that a cheese plough-
man’s sandwich was in the top ten favourite fillings in 2008 and
2009 for the UK population.
For each auction, five variants of the cheese ploughman’s sand-
wiches were made, each only varying the bread type: White (W),
Whole Grain (WG), Half White/Half Whole Grain (HH), Whole
Grain Granary (WGG) and a White bread substitute for Functional
(Inulin) bread, communicated to participants as the ‘functional’
bread product (FF). Inulin bread is only available in small experi-
mental batches and due to logistical difficulties it was not actually
used in the food auction. White bread was used as a substitute for
the functional product as they share the same taste and texture
characteristics. The nutritional information presented with the
‘functional’ bread was based on white bread but with adjusted fi-
bre content.
All sandwiches were prepared using a standard recipe weighed
by electronic scales on the morning of the auction to ensure fresh-
ness. This also minimised any variation between the sandwiches to
allow participants to be able to identify the difference as a result of
the bread type rather than the sandwich contents. Each sandwich
consisted of two slices of medium sliced bread, 6 g of vegetable
fat spread, one prepared pre-sliced mild Edam cheese slice, 10 g
of small chunk sandwich pickle and 20 g prepared iceberg lettuce.
The aim of the experiment was to determine the difference in
WTP due to bread type and health claim. The generic health claims
purely identified if the fibre content of the bread product was high-
er than the white bread. The specific whole grain health claim used
the Joint Health Claim Initiative (2002) approved claim, whilst the
specific health claim for the functional bread was provided by Pre-
mier Foods to be tested.
For this experiment a Vickrey second price auction with a full
bidding process was used. This choice of auction mechanism is
popular and relative performance compared to alternatives such
as BDM is good as demonstrated by Rozan et al. (2004). Further-
more, this auction mechanism performs well in induced and
non-induced value auctions, it is easy to explain to participants
and deals with ‘on margin’ bidders (Lusk and Shogren, 2007). In
the case of pre-packed sandwiches, they can be considered an ‘on
margin’ product, because the majority of working adults are likely
to have an appreciation of the market value, even if they do not
regularly purchase the product. Other auction techniques including
the BDM and random nth price auction are less well suited for on
margin products. Furthermore, the BDM and random nth price auc-
tions would have required an unknown number of sandwiches to
be made and held in reserve for the end of the auction when the
binding product and round are selected, increasing waste and asso-
ciated costs.
Full bidding was selected to maximise participant engagement
and interest in the auction process, and also reduce the amount
of sandwiches required per auction, minimising waste. The alter-
native to full bidding is endowment bidding, but it can introduce
bias via researchers’ choice of the initial product provided to par-
ticipants. This can influence the bids for the second product on of-
fer (e.g. change from the product provided to a different product).
Participants may be adverse to loss and unwilling to risk exchang-
ing the endowment for an alternative product (Lusk and Shogren,
2007).
We decided not to employ a reference price or reveal bids as
rounds progressed during the auction as both approaches have
been shown to effect WTP estimates. For example, Bernard and
He (2010) show that reference prices can affect bids, whereas Dric-
houtis et al. (2008) showed higher bids were received when refer-
ence price information was provided versus no information. In
addition, Corrigan and Rousu (2011) report that when comparing
a second price Vickrey auction with the BDM that with the second
price auction participants change bids in the manner predicted by
theory, but this does not happen if price feedback is employed. In-
deed references cited therein recommend that in repeated second-
price auctions that there should be no price feedback. In the auc-
tion we report there is no price feedback between rounds.
At least 10 potential participants were invited to each auction.
Unknown to the participants the auctions were divided into two
distinct experimental treatments. The difference in treatments oc-
curred during round two of the auction where participants re-
ceived nutritional information. Either, a specific or non-specific
health claim about the products being auctioned was also pro-
vided. Each time the auction was conducted it alternated between
the experimental treatments, with each participant being ran-
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
domly invited to one of the treatments. The order the sandwiches
were presented to participants varied in each auction to remove
any issues associated with a location affect. Overall, there was a
slight over attendance at the specific health information auctions
(six more participants).
Auction protocol
The same researcher conducted a total of 12 auctions over a
1-month period using the following protocol:
Step 1: On arrival all participants completed and signed a con-
sent form, and a form committing them to buy the product if they
won the auction. To provide anonymity each participant was pro-
vided with a unique ID. Each participant received £5 for taking part
in the auction. £5 was chosen because of the cost of a shop bought
sandwich generally lies between £2 and £4 depending on its filling.
If the participant won the auction they would be expected to use
this money to purchase the sandwich.
Step 2: Participants completed a Dutch Eating Behaviour Ques-
tionnaire (DEBQ) and a questionnaire about their current hunger
and mood levels using Visual Analogue Scales (VASs).
The DEBQ asks individuals a set of 33 questions to identify their
eating behaviours: ‘restrained’, ‘emotional’ and ‘external’ (van Stri-
en et al., 1986).1 Restrained eating theory combines the behavioural
consequences of emotional and external eating formed as a result of
dieting. Dieters (restrained eaters) can actually over consume calo-
ries when experiencing dis-inhibiting factors. It is a requirement of
the auction that the individual tastes the sandwich and for one par-
ticipant to purchase the sandwich. Whilst this experiment does not
easily allow the individual to over consume food, it is possible that a
person who has been identified as a ‘restrained’ eater may not place
a bid that reflects their true valuation of the products on offer, as the
dis-inhibiting factors have not been removed from the situation.
The emotional eating concept is based on psychosomatic the-
ory: an individuals‘ normal reaction to emotional states would be
to lose their appetite, however, some people respond by eating
more (van Strien et al., 1986). As the auction is still an experimen-
tal procedure and does not fully reflect a real life situation there is
the potential that participants of the auction would experience
stress or anxiety, which could potentially impact their bid to be
higher or lower than the value they really place on the product.
Externality theory addresses the issue of external eating, which
deals with an individual’s response to food stimuli regardless of
their level of satiety. In this case the individual participating in
the food auction maybe influenced by other participants around
them to over value the sandwich to ensure that they have a higher
chance of being the purchaser at the end of the experiment. There-
fore, the three concepts were recorded as control variables for use
during data analysis to explore if the concepts of restrained, emo-
tional and external eating influenced the bids placed during the
auction.
Participants’ mood and hunger was also assessed using a stan-
dard series of Visual Analogue Scales (VASs) designed by Flint
et al. (2000) and Martins et al. (2007). Participants were asked to
place a series of vertical marks on a 10 cm line between two ex-
tremes (e.g. ‘‘I am not hungry at all’’/‘‘I have never been more hun-
gry’’). VAS’s are a reliable indicator of participants’ desire to eat as a
result of how hungry the individual is and are suited for use under
1
The DEBQ was constructed from three pre-existing questionnaires (Eating
Patterns Questionnaire, The Fragenbogen für Latente Adipositas and the Eating
Behaviour Inventory), which identified 100 items that have been used to measure
eating behaviour. The questionnaire was trialled and redefined through a period of
studies to reduce the number of questions down to the 33 items it now uses. This was
achieved by using factor analysis, goodness of fit (orthogonal congruence rotation)
and internal consistency (Cronbach alpha).
235
experimental conditions (Flint et al., 2000). This research used the
hunger VAS from Flint et al. because it was easy to explain and is a
reliable predictor of participants desire to eat (and therefore possi-
bly bid higher). The mood rating employed the same format as the
hunger questionnaire following Martins et al. All the questions
asked on the VAS can be seen in Table 4.
Step 3: Participants were then fully briefed on the procedure of
the auction method using a PowerPoint presentation and script, to
ensure consistency. Within this presentation, information about
the auction technique was discussed. Participants were informed
about the dominant strategy ensuring they were able to bid their
true value for the products on offer. During this time participants
were shown examples of bidding from an auction scenario.
Step 4: A training auction was conducted using crisps and fol-
lowed the same number of rounds as the actual sandwich auction.
This auction, used a different good to train participants, allowing
participants to understand bidding behaviour without changing
the bids received for the sandwiches. As with the actual auction,
once all the bids had been collected they were placed into ascend-
ing price order. The person with the highest bid was required to pay
the second highest price. At this point participants were allowed to
talk and ask the facilitator questions about the auction procedure.
The use of the training auction meant that multiple rounds using
the product of interest was deemed unnecessary to ensure partici-
pants were familiar with the auction mechanism. Some studies em-
ploy multiple rounds within auctions to help participants
understand the mechanism. However, this learning requires feedback
about the bids received, which introduces affiliation affects and there-
fore, changes the individual’s true bid (Lusk and Shogren, 2007).
Step 5: The main auction was now undertaken. During the auc-
tion, each participant was sat separately. Participants could not
confer or talk during the actual bidding process.
The auction was split into three rounds, which followed the
same order:
Round One – Participants arrive and depending on which of the
12 auction sessions they attended they are placed in one of two
experimental treatments. In round one this selection is not re-
vealed. Individually participants were shown five pictures of the
sandwiches and asked to rate them based on appearance on a se-
ven-point likert scale. Once this information was collected each
participant had to submit a sealed bid for each of the five products.
Round Two – Both experimental treatments are now provided
with the same nutritional information about the sandwiches
(worked out using the values provided by each manufacturer and
adjusted for weight). The nutritional information was presented
to the participants as per 100 g and per sandwich (two slices of
bread with filling) and used the UK traffic light system colour cod-
ing for four key components (fat, saturated fat, sugar and salt)
(Food Standards Agency, 2010).2
The key difference in information provision is related to health
claims. One experimental treatment received a non-specific health
claim about the bread. The non-specific health claim for any bread
type containing whole grain was:
‘Naturally Rich in Whole Grains’.
and for the functional product:
‘Source of Fibre’.
In the second experimental treatment, participants received a
specific health claim about the bread. The specific health claim
for whole grain bread types was:
2
UK consumer understanding of the traffic light system of nutritional labelling is
well established in the literature, e.g., Balcombe et al. (2010).
236 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
‘People with a healthy heart tend to eat more whole grain foods as
part of a healthy lifestyle’ (Joint Health Claims Initiative, 2002).
and for the functional bread containing Inulin:
‘Inulin from chicory promotes healthy gut bacteria. This product
provide a third of the suggested 5 g/day of Inulin’ (provided by
Premier Foods).
Participants then re-rated each sandwich based on the nutri-
tional and health information provided using a seven-point likert
scale. Once this information was collected participants were re-
quired again to submit a sealed bid for each of the five products.
Round Three – Participants then tasted the five sandwiches on
offer. They then rated the attributes of the bread on a seven-point
likert scale. Once again this information was collected and then a
sealed bid was collected for each of the five products.
All participants placed a bid for each of the sandwiches on offer
during each of the three auction rounds (five sandwiches over
three rounds, totalling 15 different bids produced per participant).
If the participant did not want to purchase the sandwich on offer
they would still need to return a bid that stated zero. The bids that
participants’ submitted in the auction were sealed and no other
participant was aware of another participants bid. All bids were
collected at the end of each round prior to the commencement of
the next round.
Step 6: Once all rounds had been completed, one of the rounds
and one of the products were selected at random as ‘binding’. In
the binding round the bids received for the chosen product were
arranged in ascending order and the person who had the highest
bid had to pay the second highest price and received the sandwich.
Whilst the winning participants were not expected to consume the
sandwich during the auction they were expected to take the prod-
uct away and consume it later. The payment made by the partici-
pant was paid out of the initial endowment of £5. Therefore, the
participant would leave the auction with £5 less the second highest
price received for the sandwich that became the binding product.
This was the only time that participants had any feedback about
the bids received during the auction.
Step 7: As a result of the functional bread being substituted by
white bread, participants were told of the use of a proxy (the white
bread and the functional bread sandwich were actually the same
sandwich), during the debriefing. At this point participants were
able to ask questions about the auction as well as discuss how they
felt about the use of the proxy.
Step 8: The final step required participants, prior to leaving the
experiment, to complete a questionnaire collecting information
about their socio-demographic characteristics.
Table 1
Summary of average bids received and the average rating (by round).
White bread Whole grain bread
sandwich sandwich
Round 1 Mean bid (£) 0.87 1.11
Variance 0.49 0.71
Skewness 0.69 1.18
Round 2 Bid (£) 0.84 1.17
Variance 0.43 0.67
Skewness 0.56 0.87
Round 3 Bid (£) 0.83 1.15
Variance 0.54 0.63
Skewness 0.88 0.74
Survey data and results
Descriptive statistics
In total we recruited a total of 138 participants (53 males and
85 females) from University staff and students. The sample size
for the auction was chosen based on similar food auction experi-
ments using perishable goods (Poole et al., 2007; Rousu et al.,
2007, and Ginon et al., 2009). Participants were excluded if they
had an allergy to any of the sandwich ingredients.
The sample was composed of 62% female, with average in-
come approximately £25,000 per annum. This is a little below
average income for UK households which in 2006–2007 was
£30,000. In addition, 72% of the sample is the primary shopper
and 94% claim to have some knowledge about whole grains.
The percentage of primary shoppers is reflected in the number
of single individuals, 52%, with 47% of the sample aged between
18 and 27 years.
The age distribution of the sample reflects the inclusion of stu-
dents in the sample. This might be considered a limitation of the
study. However, there is support in the literature for similar sam-
ple composition. Depositario et al. (2009) compared student and
non-student groups WTP in an experimental auction, and found
that there little difference in bids received between the groups
and that over multiple rounds bidding behaviour converged, sup-
porting the use of students in auctions. Also, Drichoutis et al.
(2009) provide another defence when employing a student only
sample. They consider sandwich choice, which they refer to as
Food-Away-From-Home. This type of product they claim is aimed
at younger people, and as such a sample of students does not nec-
essarily bias results. However, we acknowledge that as a result of
the sample characteristics we do not claim that the sample is rep-
resentative of the UK population.
Next we present the raw bid data and the associated ratings for
each type of sandwich for rounds one and two. As previously
noted, during the auction participants were asked to provide a
bid and corresponding appeal rating (round one and two). This
information is summarised in Table 1.
From Table 1 we observe that participants’ WTP was higher for
whole grain and whole grain granary bread sandwiches. We note
that the mean bids in round one were higher for breads containing
more fibre (apart from the new bread). In round two this trend
continues, although after the nutritional information was received,
bids increased for all breads apart from white bread. In round three
once participants had been able to taste the bread, only the whole
grain granary bread bid increased.
The variance of the bids shows that, for the white and whole
grain sandwich, decreases between rounds one and two and in-
creases between rounds two and three. The half white and half
whole grain, and the functional bread sandwich shows an increase
across all three rounds. The whole grain granary bread sandwich
Half and half bread Whole grain granary Functional bread
sandwich bread sandwich sandwich
0.96 1.23 0.85
0.54 0.78 0.51
0.66 1.00 0.76
1.13 1.24 0.92
0.67 0.85 0.52
1.01 1.02 0.63
1.00 1.28 0.89
0.70 0.77 0.71
1.84 0.89 1.73
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
Fig. 1. Range of bids received during the auction rounds for the white bread
sandwich.
Fig. 2. Range of bids received during the auction rounds for the whole grain bread
sandwich.
Fig. 3. Range of bids received during the auction rounds for the half white and
whole grain bread sandwich.
bid variance initially increases (between one and two) and then de-
creases (round two and three).
Turning to the skewness of the bids, the distributions of the bids
do not follow the normal distribution, as the degree of skewness is
over twice the level of the standard error of skewness. Also, the cal-
culated skewness value changes between the products and the
rounds. For the white bread sandwich, the skewness decreases be-
tween rounds one and two but increases between rounds two and
three with the final value being larger than that observed in round
one. The whole grain sandwich value decreases across the three
rounds. For the half white and half wholegrain sandwich and
whole grain granary, skewness increases then decreases, across
the three bidding rounds. Finally, for the functional bread sand-
wich values initially decrease between the first two rounds and
then increase significantly between rounds two and three. This in-
crease can in part be explained by a number of high bids in round
three for the functional bread. Indeed, we can see this effect by
examining the distribution of the bids received during the auction.
237
Fig. 4. Range of bids received during the auction rounds for the whole grain granary
bread sandwich.
Fig. 5. Range of bids received during the auction rounds for the functional (Inulin)
bread sandwich.
Figs. 1–5 show the distribution of bids received over the three auc-
tion rounds for each of the five bread types.
From Figs. 1–5 we can see that there are a significant number of
zero bids received and the number was highest for the white and
functional (Inulin) bread sandwiches. For all bread types the bids
peak at £1.00. Higher bids were received from participants for
the half white and half whole grain sandwich and the whole grain
granary sandwich. We can also see that maximum bid for white
bread was just over £3.00, for wholegrain £4.40, for half white
and half whole grain bread £5.00, for whole grain granary £5.00
and for the functional bread £5.00.
Finally, in order to better understand how the skewness and
kurtosis change between rounds we have calculated 95% confi-
dence intervals for both measures for each bread type for each
round. To do this we employed ‘bootstrapping’, whereby we drew
with replacement samples of n = 130 and we repeated this 10,000
times. Our results are presented in Table 2.
As can be observed from Table 2 for almost all bid distributions
(by experimental treatment and for pooled data) we have statisti-
cally significant positive skewness. However, when we examine
our results for kurtosis we find that almost all confidence intervals
contain zero. Although these distribution exhibit varying degrees
of both positive kurtosis (leptokurtic) and negative kurtosis (platy-
kurtic) this measure of peakedness is generally insignificant.
Statistical methods
The main focus of the analysis we present in this paper is how
respondents react to the provision of specific and non-specific
health claims about the type of bread they eat. To identify this ef-
238 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245

Table 2
95% Confidence intervals for skew and kurtosis of bids.

Non-specific Specific All

Skewness Kurtosis Skewness Kurtosis Skewness Kurtosis
0.025 0.975 0.025 0.975 0.025 0.975 0.025 0.975 0.025 0.975 0.025 0.975
Round 1
White 0.292 1.285 1.127 1.925 0.180 0.769 1.131 0.09 0.299 0.993 0.930 0.679
WG 0.739 2.039 0.073 5.684 0.424 1.116 0.628 1.246 0.572 1.661 0.406 3.887
HH 0.329 0.941 0.914 0.397 0.208 0.854 1.042 0.389 0.301 0.976 0.967 0.783
WGG 0.468 1.996 0.665 5.797 0.108 0.855 0.818 0.815 0.265 1.562 0.819 3.995
Inulin 0.438 1.304 0.965 1.654 0.230 0.921 0.947 0.614 0.335 1.082 0.912 0.979
Round 2
White 0.038 0.663 1.298 0.364 0.107 0.891 1.048 0.728 0.097 0.928 1.159 0.919
WG 0.516 1.515 0.441 2.903 0.238 1.000 0.639 1.147 0.368 1.244 0.593 1.974
HH 0.531 1.865 0.514 4.859 0.050 0.635 0.847 0.370 0.181 1.520 0.832 3.837
WGG 0.578 1.901 0.657 4.846 0.209 1.003 0.783 1.053 0.371 1.532 0.802 3.424
Inulin 0.453 1.257 0.898 1.421 0.109 0.700 0.822 0.227 0.257 0.923 0.882 0.516
Round 3
White 0.477 1.184 0.809 1.087 0.468 1.062 0.796 0.720 0.510 1.173 0.684 1.099
WG 0.245 1.155 1.096 1.375 0.292 1.081 0.790 1.269 0.279 1.103 0.835 1.329
HH 1.395 2.612 2.444 9.672 0.281 1.030 0.640 1.144 0.480 2.338 0.422 8.481
WGG 0.240 1.788 1.195 4.675 0.074 0.924 0.626 1.263 0.116 1.473 0.970 3.746
Inulin 1.298 2.814 2.715 11.86 0.425 1.015 0.678 0.815 0.619 2.306 0.105 8.395

Note: W = White bread; WG = Whole grain; HH = Half & Half; WGG = Whole grain granary; FF = Functional bread.

fect we conduct a number of non-parametric hypothesis tests and where bidi,j,kl is the bid made by individual i for bread type j in
also estimate several regression models with our auction data. For round k and round l, x is a vector of independent variables and ei
all regression models the dependent variable is either the bid in a a normally distributed error term with mean zero and variance
specific round or the difference in bids made by an individual be- r2. Furthermore, like Hustvedt and Bernard (2008) and Bernard
tween two rounds. and Bernard (2009) it is highly likely that the data being employed
The regression analysis we conduct with our data is similar to will be subject to heteroscedasticity. So we model the variance as
that of Rousu et al. (2007) and Marette et al. (2010) in that we r2(exp(zic)) where zi is a vector of variables subject to heterosced-
examine differences in bidding behaviour between the rounds of asticity and c is vector of coefficients to be estimated. Finally, we re-
the auction so that we can isolate the effect of information provi- port marginal effects and discrete changes for all models.
sion. Specifically, our regression analysis presents three sets of re-
sults. First, we estimate a model to examine the bids made in the
Variable construction
first round of the auction (R1). Second, we run a regression to
examine the difference in bids between the first and second rounds
Our analysis employs a number of control variables for partici-
so as to see the impact of nutritional information (i.e., R2R1). Third,
pants in the auction. VAS1, VAS2, VAS3 and VAS4, and the sensory
we run a regression to explain the difference between the second
attributes collected during round three, are constructed from data
and third rounds to see how tasting the sandwiches impacts the
collected from the VAS. Due to the quantity of data collected we
bids (i.e., R3R2). Finally, we pool all the bread types so as to in-
undertook factor analysis to reduce the number of variables to be
crease the statistical power of the analysis. To take account of
included in the regression analysis. A Kaiser–Meyer–Olkin (KMO)
the different bread types we employ bread specific dummy vari-
test indicated that the VAS scores for mood and hunger were free
ables in all models.
from multicollinearity and were suitable for factor analysis (i.e.
In keeping with these earlier studies we employ a censored
KMO > 0.5). Thus, the 15 variables from the mood and hunger
regression model (i.e., a Tobit specification). As the bids obtained
VAS were reduced to a total of four factors, each with an eigen va-
in the auction are censored at zero and five we employ a double
lue greater than one. Cumulatively the four VAS factors explained
Tobit specification for all models estimated. In terms of the models
66.4% of the variance in the data collected. Cronbach alpha is a
estimated, the Tobit specification for model R1 is conventional
widely used measure of internal consistency for scale data ensur-
whereas the specification for R2R1 and R3R2 is more complicated.
ing that the questionnaire is actually measuring mood and hunger.
Like Rousu et al. (2007) and Hustvedt and Bernard (2008) we note
The threshold value for Cronbach alpha is 0.7. A Cronbach alpha of
that when we estimate models R2R1 and R3R2, because the data is
0.67, 0.18, 0.53 and 0.38 was calculated for factor one, two, three
censored in each round of the auction the resulting differences in
and four respectively for the VAS factors. As such we expected only
bids are censored as well. Thus, we assume that the latent differ-
VAS1 and VAS3 to yield any statistically significant results, and this
ence in bids between rounds (k and l) are related to the actual dif-
is borne in our analysis. A summary of VAS is presented in Table 3.
ference in the bids (Yi,j,kl) observed during the auction as follows:
In addition, we also asked all participants to rate each sandwich
8 
>
> Y i;j;kl ¼ xb þ ei if 5 > bidi;j;k > 0 and 5 > bidi;j;l > 0 using 13 sensory attributes in round three of the auction. Each of
>
> the 13 sandwiches attributes for each of the five sandwiches was
>
> ½bidi;j;k ; 1Þ if bidi;j;l ¼ 0
>
>
>
> included in a separate factor analysis. The KMO test indicated that
>
< ½bid i;j;l ; 1Þ if bidi;j;k ¼ 0
the sensory attributes for each sandwich were free from multicol-
Y i;j;kl ¼ ½bidi;j;k ; 5; 1Þ if bidi;j;l ¼ 5 linearity and were suitable for factor analysis (i.e. KMO > 0.5) with
>
>
>
> ½1; 5  bidi;j;l Þ if bidi;j;k ¼ 5 an eigen value threshold of greater than one. We reduced the set of
>
>
>
> sandwich attributes to five factors for each bread type explaining
>
> ð1; 1Þ if bidi;j;l ¼ 5 and bidi;j;k ¼ 0
>
:
ð1; 1Þ if bidi;j;l ¼ 0 and bidi;j;k ¼ 0 63% of the variance in the data. However, we have not included
these variables in our analysis, as they provided no additional sta-
ð1Þ
tistical power. A summary of these results is presented in Table 4.
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245 239

Table 3
Factor analysis results for the VAS questionnaire based on mood and hunger.

Visual analogue scale question Factor 1 Factor 2 Factor 3 Factor 4

I feel drowsy/alert 0.610
I feel tense/relaxed 0.739
I feel sad/happy 0.857
I feel angry/friendly 0.847
I feel uncertain/confident 0.794
I feel muddled/clear/headed 0.709
I feel bored/interested 0.796
How hungry do you feel? 0.836
How much do you think you can eat? 0.764
How full do you feel? 0.756
How thirsty do you feel? 0.834
Would you like to eat something sweet? 0.793
Would you like to eat something salty 0.714
Would you like to eat something savoury? 0.639
Would you like to eat something fatty? 0.567
Cronbach alpha 0.67 0.18 0.53 0.38

Table 4
Factor analysis of the sensory attributes for round three of the food auction.

Bread type Factor Sweet Salty Buttery Sour Wheaty Dense Springy Firm Moist Soft Gummy Crunchy Chewy Alpha
W 1 0.523 0.724 0.589 0.750 0.58
2 0.644 0.652 0.751 0.54
3 0.843 0.546 0.678 0.54
4 0.587 0.813 0.37
5 0.878 –
WG 1 0.547 0.573 0.632 0.49
2 0.652 0.806 0.747 0.731 0.595 0.55
3 0.767 0.690 0.50
4 0.780 0.786 0.55
5 0.790 –
HH 1 0.622 0.704 0.806 0.68
2 0.504 0.796 0.713 0.453 0.17
3 0.573 0.796 0.630 0.39
4 0.760 0.803 0.48
5 0.772 0.37
WGG 1 0.518 0.863 0.785 0.528 0.45
2 0.692 0.808 0.810 0.42
3 0.777 0.722 0.43
4 0.829 0.768 0.68
5 0.827 0.733 –
FF 1 0.483 0.768 0.715 0.48
2 0.793 0.715 0.62
3 0.770 0.303 0.639 0.612 0.47
4 0.794 0.550 0.725 0.68
5 0.677 –

Results ceived the specific nutritional information decreased their bids

To initially test if there were associations between the differ-
ences in the bids received between the rounds (i.e. the delta price)
and the bread type, non-parametric Mann–Whitney U tests were
employed to test the following hypothesis:
H0 – There is not an association between nutritional informa-
tion and delta price (change in price) between the rounds.
H1 – There is an association between nutritional information
and delta price (change in price) between the rounds.
The main results are reported in Table 5.
The Mann–Whitney U results between the different nutritional
information groups only showed an association between rounds
two and three. The statistically significant results are found for
the half white and half whole grain bread and the functional (Inu-
lin) bread sandwiches. For the half white and half whole grain
bread sandwich, the delta price showed that participants who re-
more between round two and three (after participants received
the nutritional information and were then required to taste the
sandwich), than participants who received non-specific nutritional
information. Between rounds two and three participants who re-
ceived non-specific nutritional information increased their bids
for the functional bread, participants who received the specific
nutritional information decreased their bids between round two
and three.
A weakness of the non-parametric test results is that they only
compare one variable to another. In this case there are multiple
variables that could impact an individual’s bid. Thus, we know
present multiple regression results employing the Tobit specifica-
tions. In our regressions the dependent variable is either the bid
made or the difference in the bids made between rounds. The set
explanatory variables employed are presented in Table 6.
Our regression results are presented in Table 7 for round one
(R1) and then for differences between rounds two and one
(R2R1) and rounds three and two (R3R2).
240 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245

Table 5
Mann–Whitney U results for the two nutritional information states and delta price.

Delta price Bread type Univariate analysis Mean delta price Round 2–3
Non-specific Specific
Round two and three HH 1 0.096 .0906 .1828
Functional (Inulin) bread 1 0.074 0.530 .1308

Table 6
Explanatory variables.

Variable Label Units Average

White bread W Yes = 1, 0 = No 0.2
Whole grain WG Yes = 1, 0 = No 0.2
Half and half HH Yes = 1, 0 = No 0.2
Whole grain granary WGG Yes = 1, 0 = No 0.2
Functional bread FF Yes = 1, 0 = No 0.2
Gender Gender Male = 0, Female = 1 0.62
Income Income 1 – Less than £12,000 2.63
2 – £12,000–21,999
3 – £22,000–31,999
4 – £32,000–41,999
5 – £42,000–51,999
6 – Greater than £52,000
Single SINGLE Yes = 1, 0 = No 0.47
Age AGE Years 35
VAS 1 – Mood VAS 1 Factor analysis score Min = 3.363
Max = 2.152
VAS 2 – Hunger VAS 2 Factor analysis score Min = 3.695
Max = 2.167
VAS 3 – Thirst VAS 3 Factor analysis score Min = 2.594
Max = 2.596
VAS 4 – Indulgence VAS 4 Factor analysis score Min = 3.106
Max = 5.806
DEBQ restrained eaters DEBQR 1 – Never 2.51
3 – Sometimes SD = 0.885
5 – Very often
DEBQ external eaters DEBQEXT 1 – Never 3.31
3 – SOMETIMES SD = 0.75
5 – VERY often
DEBQ emotional eaters DEBQEMO 1 – Never 2.57
3 – Sometimes SD = 0.79
5 – Very often
Primary shopper PRIMSHOP Not shopper = 0, shopper = 1 0.72
Knowledge of whole grains KNOWWG Not know = 0, know = 1 0.94
Visual appeal (Round 1) for all sandwiches LOOKSCORE 1 – Not at all appetising W = 3.80
4 – No opinion WG = 4.49
7 – Extremely appetising HH = 3.90
WGG = 4.81
FF = 3.79
Nutritional appeal (Round 2) for all sandwiches NUTSCORE 1 – Not at all appealing W = 3.08
4 – No opinion WG = 4.81
7 – Extremely appealing HH = 4.69
WGG = 5.01
FF = 3.94
Nutritional information NUTINFO Yes = 1, 0 = No 0.52

The results presented in Table 7 are composed of two parts. The
top half presents results for the relationship between bids and dif-
ference in bids and the various explanatory variables. The bottom
part shows results explaining the sources of heteroscedasticity in
the data. The variables included in the modelling of heteroscedas-
ticity are those we identified during model estimation.
Beginning with the results for round one (R1), we can see
that the dummy variables for all bread types are statistically sig-
nificant and positive. Interestingly, the functional bread has the
smallest coefficient indicating the lowest bid, which is consistent
with the raw data. The bread type with the largest coefficient is
the whole grain granary. We can also see that as we would ex-
pect, that the score rating given by respondents to how a partic-
ular sandwich appears (LOOKSCORE) is also positively related to
the bid.
In terms of the socio-economic variables older individuals bid
slightly less than younger participants, and that single people bid
more. We also see that if a participant is the primary food shopper
it yields a higher bid, but prior knowledge of whole grains is not
statistically significant.
If we consider the various constructs none of the VAS constructs
are statistically significant. However, both the DEBQR, which is
negative and DEBQEMO, which is positive are statistically signifi-
cant. The negative coefficient for DEBQR indicates that potentially
some participants have not place a bid that reflects their true val-
uation of the products on offer as dis-inhibiting factors have not
been controlled for. In contrast, the positive coefficient on DEBQ-
EMO suggests that some participants may have experienced some
degree of stress during the experimental auction that positively
influenced the bids they have made.
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245 241

Table 7
Tobit regression results.

Variable R1 R2R1 R3R2

Marginal P value Marginal P value Marginal P value
W 0.606*** 0.003 0.087** 0.012 0.049 0.552
WG 0.766*** 0.000 0.109*** 0.002 0.064 0.423
HH 0.693*** 0.001 0.157*** 0.000 0.092 0.247
WGG 0.860*** 0.000 0.079** 0.036 0.006 0.937
FF 0.597*** 0.003 0.119*** 0.000 0.083 0.276
AGE 0.001*** 0.000 0.0001 0.371 0.0001 0.475
INCOME 0.012 0.413 0.009* 0.066 0.014*** 0.000
SINGLE 0.145** 0.012 0.018 0.320 0.055** 0.026
PRIMSHOP 0.176*** 0.006 0.06*** 0.001 0.080*** 0.001
KNOWWG 0.024 0.887 0.15*** 0.000 0.079 0.285
VAS1 0.024 0.399 0.02*** 0.007 0.007 0.494
VAS2 0.042 0.121 0.006 0.412 0.007 0.461
VAS3 0.023 0.336 0.021*** 0.004 0.009 0.314
VAS4 0.005 0.865 0.004 0.606 0.004 0.638
DEBQR 0.084** 0.024 0.006 0.607 0.017 0.166
DEBQEXT 0.057 0.125 0.015 0.191 0.025 0.146
DEBQEMO 0.140*** 0.001 0.010 0.430 0.008 0.679
LOOKSCORE 0.124*** 0.000 0.02*** 0.000
NUTSCORE 0.038*** 0.000 0.003 0.592
NUTINFO 0.026 0.147 0.016 0.448
Heteroscedasticity
AGE 0.0001** 0.021 0.0001 0.688 0.001** 0.014
INCOME 0.006** 0.021 0.003 0.492 0.139*** 0.000
KNOWWG 0.049** 0.019 0.139*** 0.000 0.579*** 0.000
DEBQR 0.011*** 0.001 0.003 0.740 0.138*** 0.000
DEBQEXT 0.019*** 0.000 0.005 0.590 0.024 0.648
DEBQEMO 0.008 0.179 0.002 0.885 0.162*** 0.004
Sigma 1.194*** 0.000 0.262*** 0.000 1.074*** 0.000
*
Statistically significant at 10% levels.
**
Statistically significant at 5%.
***
Statistically significant at 1%.

Finally, for R1 we can examine the sources of heteroscedasticity
in the data. Both age and income have positive coefficients, albeit
very small. In contrast knowledge about whole grains, had a nega-
tive impact, which implies that those with more knowledge bid
less than those without prior knowledge of whole grains. Also DEB-
QR and DEBQEXT are identified as sources of heteroscedasticity, al-
beit in different directions.
Next we consider the difference in bids between rounds two
and one (R2R1). The first things to note is that the dummy vari-
ables for the bread type are all positive and all statistically signif-
icant. This indicates that the provision of information (ie
nutritional and health) has had a positive impact on the bids. We
can also see that the smallest effect is for whole grain granary
and white bread, whereas the effect on the functional bread is
somewhat larger. Thus, the effect of the health information as it re-
lates to the functional bread has increased the magnitude of bids.
This result is in keeping with Bitzios et al. (2011) who found that
consumers are WTP more for a product that has an explicit health
claim as opposed to a functional ingredient. Indeed, when asked,
participants stated that they preferred the simple non-specific
health claim for the new functional bread rather than the specific
health claim. We also found that most participants had not heard
of the functional ingredient, Inulin, with some confusing it with
Insulin, which then put them off the product. Clearly, further work
will be needed to improve consumers understanding of the poten-
tial benefits of Inulin in the diet if products such as enhanced white
bread are to be credible. This may not be an easy task.
Next we can compare how survey participants difference in bids
relate to the subjective scores provided for how the sandwiches
looked in Round 1 (LOOKSCORE) compared to the nutritional infor-
mation provided (NUTSCORE) in Round 2. We can see that the way
the bread was scored in round one in terms of the way it looked is
negatively related to the bid difference. Thus, the provision of
information has changed participants’ valuations of the bread,
and as we would expect the score reported in round two is posi-
tively related to the bid difference.
Importantly, we observe that the coefficient for the difference in
experimental treatment relating to health information (NUTINFO)
is positive but not statistically significant. This result indicates that
differences in the provision of health information (specific or non-
specific) have not had an effect on the difference in the bids made.
This result is important as we have seen an increase in bids for all
bread types, but we are unable to attribute this to the experimental
treatment offering the specific health claim. In addition, we also
find that the provision of the information in round two did not
have the same impact on individuals who had prior knowledge
of whole grains. This coefficient is negative and statistically
significant.
There are several other differences between the results for mod-
els R1 and R2R1. Income is now significant, albeit small and nega-
tive and also being the primary shopper has a negative impact.
These results imply that being wealthier and being the main shop-
per reduce the impact of health information on the bid. We also ob-
serve that some of the VAS coefficients are now statistically
significant. Specifically, VAS 1 is negative and significant identify-
ing that mood had an association with bids. VAS 3 is positive and
significant showing that their reported fluid needs has an impact
on their bid.
Finally, for R2R1 we can examine the sources of heteroscedas-
ticity. With this specification only prior knowledge of whole grains
has an effect. However, this positive coefficient is relatively large
indicating a significant degree of heterogeneity in the bids made
in round two from participants who claim to have a prior knowl-
edge of whole grains.
We now examine the final model, which estimated the differ-
ence in bids between rounds two and three (R3R2). The results
242 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245

confirm that the main statistical differences in the bids occurred
between rounds one and two. There is little statistical significance
between these rounds. Indeed, it would appear that being single,
the primary shopper and having prior knowledge of whole grains
had a negative impact on the difference in bids between rounds.
Thus, these results confirm that tasting the sandwiches did not
have a strong impact on the bids. However, what we observe is
a significant amount of heteroscedasticity in this model reflected
in the number and magnitude of the parameter estimates. This
indicates that far more heterogeneity in response, reflected in
the bids made, results from tasting the sandwiches compared to
the greater uniformity in response as a result of the provision of
information in round two compared to round one.
Sub-sample results
In order to examine the impact of the composition of the sample on
bidding behaviour we re-examined our data looking at specific sub-
samples.Tothiswesplitoursampleintoyoungandold(lessthanorequal
to26 years,andgreaterthan27 years).Thisyieldedsubsamplesofsize
270and 420 (approximately40:60).Theyoungsampleiscomposed of
undergraduateandpostgraduatestudents.
Like Depositario et al. (2009) we first consider differences in
bidding behaviour between groups employing a T-test. We find
no difference in average bids between groups for all bread types
for all rounds of the auction. This is a stronger result in support
of pooling the data than that presented by Depositario et al. who
rejected the null hypothesis of equal average bids between groups.
Secondly, we compare results for each of the regression models
previously reported. Our results are presented in Table 8.
The results in Table 8 reveal some interesting differences. Over-
all the results are in reasonable correspondence. However, there
are several interesting differences that are of interest. First, if we
concentrate on the estimates for the bread types we observe that
rank and magnitude are reasonably similar for models R1 Old
and R1 Young. In terms of the difference models R2R1 and R3R2
there are some interesting differences. For R2R1 the statistical sig-
nificance of the estimates is much greater for R2R1 Young. This re-
sult is supported by the fact the young cohort have reacted
positively to the difference in information treatments as signified
by the positive coefficient on NUTINFO. This means that the treat-
ment information had a stronger and statistically significant im-
pact on the younger cohort. This result is probably to be
expected as this group has a lower coefficient for the prior knowl-
edge of whole grains.
There are also differences for the R3R2 specification, with the
old cohort having all negative estimates, whereas those for young
are positive. Furthermore, some of the estimates for the young are
statistically significant (i.e., Whole grain and whole grain granary),
which is in contrast to the results reported in Table 7 for the full
sample. This suggests that the young have also responded posi-
tively to the taste of the sandwiches, which is clearly not the case
for the old. Thus, what appears to have happened at the sub-sam-
ple level is that the experimental treatment has had a stronger ef-
fect on the younger participants in the sample compared to the old.
This in part can be traced back to the prior knowledge of whole
grains, which can be seen in terms of the parameter estimates
for KNOWWG. What we observe for R3R2 is that for the young
group this has a negative impact on the difference in the bid, but
not enough to outweigh the effect of taste.
Discussion and conclusions
In this paper we have conducted an experimental food auction
to examine consumer responses to the provision of nutritional
information for a variety of bread products including a novel func-

Table 8
Sub-sample tobit regression results.

Variable R1 Old R1 Young R2R1 Old R2R1 Young R3R2 Old R3R2 Young
Marginal P Marginal P Marginal P Marginal P Marginal P Marginal P
W 0.816** 0.014 0.642** 0.041 0.071 0.291 0.033 0.444 0.105 0.239 0.077 0.150
WG 0.996*** 0.003 0.878*** 0.005 0.052 0.423 0.125** 0.012 0.099 0.262 0.087* 0.088
HH 0.921*** 0.006 0.773** 0.014 0.114* 0.081 0.129** 0.015 0.159* 0.060 0.040 0.469
WGG 1.059*** 0.002 0.948*** 0.003 0.041 0.530 0.060 0.280 0.075 0.397 0.136** 0.024
FF 0.80** 0.014 0.691** 0.030 0.083 0.191 0.078* 0.078 0.139 0.105 0.066 0.191
AGE 0.007** 0.028 0.001*** 0.000 0.001 0.174 0.001 0.213 0.001 0.274 0.000 0.970
LINCOME 0.004 0.866 0.042** 0.033 0.000 0.963 0.007 0.525 0.008*** 0.008 0.026*** 0.007
SINGLE 0.293*** 0.003 0.036 0.792 0.030 0.261 0.041 0.353 0.037 0.177 0.014 0.816
PRIMSHOP 0.116 0.353 0.099 0.295 0.089*** 0.004 0.031 0.300 0.060** 0.014 0.057 0.139
KNOWWG 0.335 0.235 0.085 0.776 0.176*** 0.000 0.127*** 0.001 0.070 0.399 0.100** 0.024
VAS1 0.126** 0.015 0.146*** 0.006 0.000 1.000 0.037** 0.030 0.001 0.990 0.008 0.698
VAS2 0.007 0.862 0.059 0.111 0.005 0.563 0.035* 0.083 0.004 0.655 0.025 0.332
VAS3 0.069** 0.043 0.032 0.558 0.022*** 0.009 0.020 0.163 0.008 0.456 0.002 0.939
VAS4 0.064 0.119 0.054 0.203 0.018** 0.042 0.019 0.264 0.013 0.243 0.018 0.355
DEBQR 0.058 0.338 0.156** 0.048 0.003 0.845 0.006 0.732 0.002 0.908 0.016 0.493
DEBQEXT 0.084* 0.072 0.013 0.880 0.008 0.584 0.036* 0.081 0.004 0.824 0.013 0.613
DEBQEMO 0.141** 0.030 0.142** 0.018 0.011 0.567 0.029 0.123 0.002 0.919 0.003 0.904
LOOKSCORE 0.126*** 0.000 0.091*** 0.000 0.021*** 0.002 0.019* 0.069
NUTSCORE 0.034*** 0.000 0.036*** 0.001 0.003 0.571 0.026** 0.046
NUTINFO 0.033 0.169 0.054* 0.079 0.026 0.258 0.034 0.381
Heteroscedasticity
AGE 0.001 0.130 0.001*** 0.001 0.002** 0.025 0.001 0.487 0.004*** 0.000 0.001 0.412
LINCOME 0.001 0.825 0.004 0.281 0.015* 0.063 0.001 0.898 0.046*** 0.000 0.016* 0.100
KNOWWG 0.064* 0.096 0.016 0.560 0.014 0.824 0.123** 0.012 0.288*** 0.000 0.259*** 0.002
DEBQR 0.001 0.889 0.013*** 0.002 0.002 0.895 0.009 0.393 0.009 0.544 0.044*** 0.001
DEBQEXT 0.027*** 0.001 0.016** 0.011 0.049*** 0.000 0.026 0.153 0.059*** 0.003 0.002 0.891
DEBQEMO 0.028*** 0.005 0.003 0.647 0.048*** 0.009 0.036 0.115 0.069*** 0.007 0.041** 0.012
*
Statistically significant at 10% levels.
**
Statistically significant at 5%.
***
Statistically significant at 1%.
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
tional white bread. Overall, this study has found that the provision
of a specific or non-specific health claim along with nutritional
information influenced participants WTP. We also found no addi-
tional impact as a result of differences in health information provi-
sion for the full sample. But, when we re-examined our data for
two sub-samples we found that younger participants did react pos-
itively to the specific health claim. This finding is consistent with
the literature in that younger participants tended to have less prior
knowledge about the health benefits of whole grains and as such
are more open to information provision. This is an important find-
ing especially as the change in information provision in our exper-
imental auction is relatively subtle. Indeed, it is potentially
tempting to employ starkly contrasting information in an experi-
mental auction so that a strong response by participants can be
identified. However, the changes made to food in the market place
tend to be more subtle and as such it is important to mimic such
changes in the auction environment. It has been identified that
consumers prefer succinct non-complex wordings for health
claims as consumers cannot in many contexts distinguish between
different types of claims made about food products (Williams,
2005). The finding that providing a health claim increases WTP is
supported by Marette et al. (2010), but not by Mialon et al.
(2002). Mialon et al. attributed this result to the fact that most
products in their study were already perceived as being high in fi-
bre so additional information did not influence the acceptance
ratings.
Our results show that in round one, respondents’ bids did re-
spond to the appearance attributes of the sandwiches. This finding
is similar to Mialon et al. (2002) who found that product colour and
the provision of the fibre content dominated flavour perception.
Then in round two the provision of nutritional information with
a specific or non-specific health claim, a form of credence attribute,
did have a strong impact on the bids. A credence attribute is an
attribute, which even after consumption a consumer is unable to
verify without the help of some external information. It is thought
that as a result of providing information, sensory perceptions of the
consumer have been altered. This is possibly as a result of a change
in consumers’ expectations of the product’s sensory properties and
sensory ratings associated with the fibre content (Mialon et al.,
2002).
A potential mechanism for the failure of most bids to increase in
round three could be due to consumer expectations not being met
whilst tasting the product after receiving the health information
(Mela, 2007). Thus, the most important attributes giving rise to a
change in bids was health information. This result is important
especially when we consider the sub-samples, where the quality
and scientific precision of the information does make a significant
difference for less well informed consumers. Overall, our results
suggest that consumers react positively to health information,
but not all appear to place a premium on specific information.
Interestingly, consumers who had prior knowledge about the
health benefits associated with a diet rich in whole grains re-
sponded differently during round two of the auction. The increase
in WTP of these ‘‘knowledgeable’’ consumers was significantly less
in round two. This suggests that their initial round one bid incorpo-
rated their pre-existing knowledge of whole grains. The provision
of additional nutritional information may only serve to confirm
the preferences of these ‘‘knowledgeable’’ consumers. These find-
ings are in keeping with earlier results reported in the literature
by Teisl et al. (2001) and Lusk et al. (2004). Within the auction
the prior knowledge of consumers was self determined and future
work may seek to identify the sources of this knowledge and if it
corresponds to publically available information.
Turning to the differences in responses between bread types,
these were observed in the bids received between the whole grain
granary bread in comparison to the functional (Inulin) sandwich.
243
This suggests participants prefer whole grain bread as a food to
provide health benefits and subsequently a source of a functional
ingredient. Thus, despite the ability of bakers to make a healthier
white bread, consumers did not appear willing to pay more for this
product compared to bread that already contains whole grain. Prior
research has shown that consumers are well aware of the health
benefits associated with whole grain foods and as such may not
view the consumption of an alternative functional product as a
necessary means of avoiding disease (Arvola et al., 2007). However,
participants mean bids were slightly higher for the functional
white bread than the traditional white bread after the health infor-
mation was presented in round two, but both breads received low-
er bids than the whole grain bread varieties.
In general, within the auction it is clear that there was a stron-
ger preference for the whole grain bread varieties than the new
bread. There was some evidence of support for the functional
white bread compared to the traditional white bread, more so after
the provision of health information. This indicates that there is bet-
ter recognition for whole grain products and their associated ben-
efits compared to more novel forms of functional ingredients. It is
also important to understand that in recent years the range of
products in the UK bread market has exploded, with each manufac-
turer offering several varieties including whole grain and diet
products. New product development has focused on the benefits
of whole grains including oats, which includes beta-glucan, an-
other possible functional ingredient and how it can help maintain
normal cholesterol levels (Hovis, 2010).
It was surprising to find that the majority of participants in the
food auction identified a preference for the whole grain bread vari-
eties due to the population they were recruited from. Participants
in the food auction were predominantly drawn from a student
population, with 47% of the sample being 18–27 year olds. Re-
search has shown that consumers of whole grain products are typ-
ically older (Lang and Jebb, 2003). Within the remit of the sample it
seems that there may be changing attitudes to whole grains. In-
deed, this can be seen by the statistically significant difference in
bid between round two and three for whole grain and whole grain
granary, for the young cohort. These results suggest the need for
more research into the dietary habits of this group.
Overall, these results indicate that more research is needed to
identify how consumers perceive the functional bread compared
to commercially available products. This may help industry to
identify if some other attributes in addition to the health benefits
associated with a functional ingredient are needed to better mar-
ket the product. There has already been a lot of research into the
acceptance of functional foods by consumers that can be called
on to effectively market a new bread product. This research has
shown that if consumers believe in the benefits of a functional food
they are more likely to accept it.
Turning to the implementation of our experimental auction
there are a number of issues that warrant further consideration.
Within the existing literature, food auctions have generally at-
tempted to control for mood and hunger, and their impact upon
the bids received. As this auction was conducted during the typ-
ical lunch period it was felt that mood and hunger were impor-
tant factors to control. The mood and hunger scores were
statistically significant in our auction, especially in round two,
suggesting that as the auction progressed it was important to
control for participants’ current hunger and mood levels. We also
note that there are a number of ways in which our auction could
be modified in the future. It was felt that the sandwiches them-
selves could be improved removing a possible source of bias in
the data. During the auction only cheese sandwiches were avail-
able for purchase. Whilst this removed a confounding factor it did
not allow all participants to express their preference for their pre-
ferred sandwich. The experimental results might have been dif-
244 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245
ferent if each participant was bidding for their preferred sand-
wich filling. This might have provided more insight into prefer-
ences towards the overall sandwich and would have lent itself
towards an endowment method for establishing WTP for a partic-
ular bread type.
Also, for consistency participants were shown photographs of
the sandwiches but this perhaps removed certain visual, tactile
and olfactory clues that might have guided participants WTP. How-
ever, the WTP estimates recorded were credible. Although the aim
of this auction was not to identify the market value of the sand-
wiches being offered, the market rate in the Vickrey auction is dic-
tated by the second highest bid received. In all auctions conducted
85% of the market rate bid was within the price range of an equiv-
alent shop brought sandwich of £1.50 and £3.50.
The sample itself maybe a limitation to the study as a high
proportion of the population were students (65 out of the 138
participants). But, the population used within the experiment
can be considered similar to that of the existing purchaser of
sandwiches. A similar observation has previously been made by
Drichoutis et al. (2009) in defence of samples with a high propor-
tion of young individuals. The market for sandwiches in 2010
showed that those aged between 25 and 54 years old account
for over 53% of all commercial sandwich sales in value terms
(The British Sandwich Association, 2010) and made up 60% of
the sample in this auction. But, it is also possible that all partic-
ipants (particularly students who are living on a low income)
may have entered into the auction with the view of maximising
their own personal gain and leaving the auction with the en-
dowed £5 rather than having to return some to the experimenter.
However, Lusk and Shogren (2007) note that the endowment
should be close to the value of auctioned good so as to allow par-
ticipants to reveal their true WTP.
Finally, in research by Roosen et al. (2007) into the effect of
health information on liking and choice, used an experiment which
endowed women with canned fish (tuna or sardines). In this exper-
iment participants initially tasted the fish before being given nutri-
tional information about the fish, at which point they got to try the
fish again. Perceptions of the fish were more positive to the health-
ier choice (sardines) even though they had stated that their initial
taste preference was for the tuna. Roosen et al. suggest that this
was because of preferences being altered from taste to a more ra-
tional decision of health information. In our auction, all partici-
pants received nutritional information ahead of tasting the
product. Whilst this procedure was validated by the focus group
it does not allow for the capturing of hedonic and impulsive bid-
ding behaviour due to taste alone. Further research might like to
vary the order of the auction rounds.
References
Anderson, J.W., 2003. Whole grains protect against atherosclerotic cardiovascular
disease. Proceedings of the Nutrition Society 62, 135–142.
Annett, L.E., Muralidharan, V., Boxall, P.C., Cash, S.B., Wismer, W.V., 2008. Influence
of health and environmental information on hedonic evaluation of organic and
conventional bread. Journal of Food Science 73 (4), 50–57.
Arvola, A., Lähteenmäki, L., Dean, M., Vassallo, M., Winkelmann, M., Claupein, E.,
Saba, A., Shepherd, R., 2007. Consumers’ beliefs about whole and refined grain
products in the UK, Italy and Finland. Journal of Cereal Science 46, 197–206.
Balcombe, K.G., Fraser, I.M., Di Falco, S., 2010. Traffic lights and food choice: a choice
experiment examining the relationship between food labels and price. Food
Policy 35, 211–220.
Bernard, J.C., Bernard, D.J., 2009. What is it about organic milk? An experimental
analysis. American Journal of Agricultural Economics 91 (3), 826–836.
Bernard, J.C., He, N., 2010. Confounded by the field: bidding in food auctions when
field prices are increasing. Agricultural and Resource Economics Review 39,
275–287.
Bitzios, M., Fraser, I., Haddock-Fraser, J., 2011. Functional ingredients and food
choice: results from a dual-mode study employing means-end-chain analysis
and a choice experiment. Food Policy 36, 714–724.
British Sandwich Association, 2009, 2010. Britain’s Favourite Sandwiches. <http://
www.sandwich.org.uk/>.
Corrigan, J.R., Rousu, M.C., 2011. Are experimental auctions demand revealing when
values are affiliated? American Journal of Agricultural Economics 93 (2), 514–
520.
Cox, D.N., Evans, G., Lease, H.J., 2007. Predictors of Australian consumers’ intentions
to consume conventional and novel sources of long-chain omega-3 fatty acids.
Public Health Nutrition 11 (1), 8–16.
Depositario, D., Nayga Jr., R., Wu, X., Laude, T., 2009. Should students be used as
subjects in experimental auctions? Economics Letters 102, 122–124.
zDewettinck, K., Van Bockstaele, F., Kühne, B., Van De Walle, D., Courtens, T., Gellynck,
X., 2008. Nutritional value of bread: influence of processing, food interaction and
consumer perception. Journal of Cereal Science 48, 243–257.
Drichoutis, A., Lazaridis, P., Nayga Jr., R., 2008. The role of reference prices in
experimental auctions. Economics Letters 99, 446–448.
Drichoutis, A.C., Lazaridis, P., Nayga Jr., R.M., 2009. Would consumers value food-
away-from-home products with nutritional labels. Agribusiness 25 (4), 550–
575.
Flint, A., Raben, A., Blundell, J., Astrup, A., 2000. Reproducibility, power and validity
of visual analogue scales in assessment of appetite sensations in single test
meal studies. International Journal of Obesity 24, 38–48.
Food Standards Agency, 2010. Traffic Light Labelling. <http://www.food.gov.uk/>.
Ginon, E., Loheac, Y., Martin, C., Combris, P., Issanchou, S., 2009. Effects of fibre
information on consumer willingness to pay for French baguettes. Food Quality
and Preference 20, 343–352.
Golan, E., Unnevehr, L., 2008. Food product composition, consumer health, and
public policy: introduction and overview of special section. Food Policy 33,
465–469.
Hellyer, N.E., Haddock-Fraser, J., 2011. Reporting diet-related health issues through
newspapers: portrayal of cardiovascular disease and Type 2 diabetes. Health
Education Research 26 (1), 13–25.
Hoare, J., Henderson, L., Bates, C.J., Prentice, A., Birch, M., Swan, G., Farron, M., 2004.
The National Diet and Nutrition Survey: Adults Aged 19 to 64 years, vol. 5,
Summary Report. The Stationery Office, London.
Hobbs, J., Sanderson, K., Haghiri, M., 2006. Evaluating willingness-to-pay for bison
attributes: an experimental auction approach. Canadian Journal of Agricultural
Economics 54, 269–287.
Hovis, 2010. Hearty Oats. <http://www.hovisheartyoats.co.uk/>.
Hustvedt, G., Bernard, J.C., 2008. Consumer willingness to pay for sustainable
apparel: the influence of labelling for fibre origin and production methods.
International Journal of Consumer Studies 32, 491–498.
Ippolito, P.M., Mathios, A.D., 1990. Information, advertising and health choices: a
study of the ceral market. RAND Journal of Economics 21 (3), 459–480.
Joint Health Claims Initiative, 2002. Generic Claims for Whole Grain. <http://
www.jhci.org.uk/>.
Kihlberg, I., Johansson, L., Langsrud, O., Risvik, E., 2005. Effects of information on
liking of bread. Food Quality and Preference 16, 25–35.
Lang, R., Jebb, S.A., 2003. Who consumes whole grains and how much? Proceedings
of the Nutrition Society 62, 123–127.
Lusk, J., Shogren, J.F., 2007. Experimental Auctions: Methods and Applications in
Economic and Marketing Research. Cambridge University Press.
Lusk, J.L., House, L.O., Valli, C., Jaeger, S.R., Moore, M., Morrow, J.L., Traill, W.B., 2004.
Effect of information about benefits of biotechnology on consumer acceptance
of genetically modified food: evidence from experimental auctions in the
United States, England and France. European Review of Agricultural Economics
31 (2), 179–204.
Mancino, L., Kuchler, F., Leibtag, E., 2008. Getting consumers to eat more whole-
grains: the role of policy, information, and food manufacturers. Food Policy 33,
489–496.
Marette, S., Roosen, J., Blanchemanche, S., Feinblatt-Mélèze, E., 2010. Functional
food, uncertainty and consumers’ choices: a lab experiment with enriched
yoghurts for lowering cholesterol. Food Policy 35, 419–428.
Martins, C., Truby, H., Morgan, L.M., 2007. Short-term appetite control in response
to a 6-week exercise programme in sedentary volunteers. British Journal of
Nutrition 98, 834–842.
Mela, D.J., 2007. Food choice and intake: the human factor. Proceedings of the
Nutrition Society 58, 513–521.
Mialon, V.S., Clark, M.R., Leppard, P.I., Cox, D.N., 2002. The effect of dietary fibre
information on consumer responses to breads and ‘English’ muffins: a cross-
cultural study. Food Quality and Preference 13, 1–12.
Mozaffarian, D., Wilson, P.W.F., Kannel, W.B., 2008. Beyond established and novel
risk factors: lifestyle risk factors for cardiovascular disease. Circulation 117,
3031.
Naylor, R., Droms, C., Haws, K., 2009. Eating with a purpose: consumer response to
functional food health claims in conflicting versus complementary information
environments. Journal of Public Policy & Marketing 28, 221–233.
Poole, N.D., Martinez, L.M.C., Giménez, F.V., 2007. Quality perceptions under
evolving information conditions: implications for diet, health and consumer
satisfaction. Food Policy 32, 175–188.
Roosen, J., Marette, S., Blanchemanche, S., Verger, P., 2007. The effect of product
health information on liking and choice. Food Quality and Preference 18, 759–
770.
Rousu, M., Huffman, W., Shogren, J., Tegene, A., 2007. Effects and value of verifiable
information in a controversial market: evidence from lab auctions of genetically
modified food. Economic Inquiry 45, 409–432.
Rozan, A., Stenger, A., Willinger, M., 2004. Willingness-to-pay for food safety: an
experimental investigation of quality certification on bidding behaviour.
European Review of Agricultural Economics 31, 409–425.
 N.E. Hellyer et al. / Food Policy 37 (2012) 232–245 245

Sidhu, J.S., Kabir, Y., Huffman, F.G., 2007. Functional foods from cereal grains.
International Journal of Food Properties 10, 231–244.
Siró, I., Kápolna, E., Kápolna, B., Lugasi, A., 2008. Functional food. Product development,
marketing and consumer acceptance – a review. Appetite 51, 456–467.
Teisl, M.F., Bockstael, N.E., Levy, A., 2001. Measuring the welfare effects of nutrition
information. American Journal of Agricultural Economics 83 (1), 133–149.
van Strien, T., Frijters, J.E.R., Bergers, G.P.A., Defares, P.B., 1986. The Dutch Eating
Behavior Questionnaire (DEBQ) for assessment of restrained, emotional, and
external eating behavior. International Journal of Eating Disorders 5, 295–
315.
Williams, P., 2005. Consumer understanding and use of health claims for foods.
Nutrition Reviews 63, 256–264.
Williamson, A.R., Hunt, A.E., Pope, J.F., Tolman, N.M., 2000. Recommendations of
dieticians for overcoming barriers to dietary adherence in individuals with
diabetes. The Diabetes Educator 26, 272–278.