2.3 Ambiguity in communicating health risk information
Two dimensions are usually considered to determine a choice situation. The first one is the relative
desirability of the possible pay-off, and the other one is the likelihood of the events that are affecting
them. The third factor that could be added is the information somebody has about the relative likelihood
of events[11]. The ambiguity of this information is “[...] a quality depending on the amount, type,
reliability and “unanimity” of information, and gives rise to one’s degree of “confidence” in an estimate
of relative likelihoods” ([11], p. 657). Camerer and Weber (1992) applied the following definition of
ambiguity: “Ambiguity is uncertainty about probability, created by missing information that is relevant
and could be known” ([12], p. 330).
Fox and Tversky (1995) argued that when people compare two events with having different levels of
knowledge about them, then the less familiar bet is less attractive compared to the more familiar one. This
is called the comparative ignorance hypothesis. That is, ambiguity aversion is assumed to be present when
subjects evaluate clear and vague prospects jointly (within-subject design), but diminishes or disappears
when the prospects are evaluated in isolation (between-subject design). The hypothesis predicts that the
clear bet will be priced above the vague bet. This discrepancy is likely to be more pronounced when clear
and vague bets are traded jointly than separately[13]. For example, Chow and Sarin (2001) showed in their
experiments that the clear bet is priced higher than the vague bet under both comparative and non-
comparative conditions [14]. In our study, we hypothesize that ambiguity in risk information influences
WTP. The ambiguity is represented by the unclear incidence rate (i.e., unclear probability of occurrence
of an E. sakazakii infection). The next section discusses the experimental design and treatments used to
test ambiguity and safe-handling information effects.
3 Experimental design
In November and December 2005, 84 mothers and fathers participated in our experiments using
Vickrey auction in a member state of the European Union (i.e., Germany). Participants were randomly
recruited either through flyer or personal communication. We were seeking parents who feed/ fed their
newborns powdered infant milk formula and are responsible for purchasing the formula. During the
recruitment, the participants were not provided information about the details of the study to avoid
participation bias related to food safety aspects of powdered infant milk formula. The subjects were
randomly assigned to one of three treatments discussed below. We conducted a total of eight
experimental auction sessions with group sizes ranging from 6 to 14 participants. Prior to the actual
experimental auction sessions, the respondents were asked to fill in an entry questionnaire containing
questions about the milk formula they feed, information sources that they use concerning baby food,
reasons for not breastfeeding, socio-economic questions and others.
3.1 Design to test ambiguity effects
The experiment was programmed and conducted with the software z-Tree [15] and involved three
treatments. The first two treatments were designed to test between-sample ambiguity effects while the
third treatment was designed to test within-sample ambiguity effects.
In Treatment 1, the participants received information about the pathogen but were not provided clear
information about the incidence rate (called “Unclear” treatment). The information about the pathogen
included information on the microorganism E. sakazakii, the diseases, symptoms and adverse health
effects it might cause, the population at risk, and the possibility that it can be found in powdered infant
milk formula. In treatment 2, the participants also received the same information about the pathogen but
unlike Treatment 1, they received clear or unambiguous information about the incidence rate (called
“Clear” treatment). The unambiguous incidence rate mentioned was one child out of 100,000 under 1 year
of age. Participants were thus asked to avoid a risk with known outcome (i.e. the symptoms) but known or
unknown likelihood of occurrence (i.e. the incidence rate), respectively. The auctions for these two
treatments to test between-sample ambiguity effects involved 5 trials each. Treatment 3 (called “Both”
treatment) (see Table 1), designed to test within-sample ambiguity effects, involved two sets of 5 trials
each. In the first set of trials, the clear or unambiguous incidence rate was not mentioned to the
participants while in the second set of trials, the participants were informed of the unambiguous incidence