General results

Prices and choices: are the two decisions consistent?
An experimental study of organic farming.
S. Blondel and M. Javaheri
Research Laboratory in Management and Economics
Institut national d’horticulture
2 rue André Lenôtre 49045 Angers cedex 01 France
[email protected]
Abstract:
The aim of this study is to establish if the preference reversal phenomenon, i.e. an
inconsistency between buying (selling) choice and WTP (WTA), can be observed in the case
of food safety valuations. We conduct an experiment with organic and conventional apples
and orange juices. There were two groups: buying decisions and selling decisions. The
average rate of inconsistency (26%) is significant, but we observe no systematic direction for
the inconsistent choices. More, there is no significant difference between the rate of
inconsistency for the buying group and the selling group.
Keywords: experimental economics, willingness to pay, organic farming,
JEL: C91, D80
1
Introduction
The modern economic theory explains the rational behavior with a set of axioms that imply
the consistency of choices. However, the observed behavior seems to violate systematically
some axioms. Since the middle of 20th century, the economists and the psychologists used
laboratory experiments to challenge the rational theory of choices. For example Herbert
Simon, Maurice Allais or the psychologists Kahneman, Tversky and Slovic have identified
various systematic deviations from the theoretical predictions.
The main purpose of this paper is to study whether some choice paradoxes can be observed
also in the case of food choices. Since many public decisions, cost- benefit analysis and
marketing researches are based on the consumer preferences, it is important to know how the
methods and techniques used to elicit preferences can affect the results.
We use the laboratory experiments for this objective. We study two well-known phenomena:
the preference reversals and the endowment effect.
The preference reversal phenomenon has been found by psychologists at the beginning of the
70s, and confirmed later with a large number of experimental results (see for instance the
surveys of Slovic, 2000, or Shogren, 2002). Most of these experiments use two types of
gambles: a safer one (smaller payoff with a higher probability) and a riskier one (greater
payoff with a lower probability). (Lichtenstein and Slovic, 1971 ; Lindman ,1971). The
preference reversal phenomenon arises when the valuations and the choice do not represent
the same order: the chosen lottery is not the one for which the individual has indicated a
higher certainty equivalent. More, the technique used for evaluations (maximum buying price
or minimum selling price) seems to affect the direction of the preference reversals (Casey
1991, 1994).
Rational choice theory suggests that with small income effects and many available substitutes,
the willingness to pay for a good and the willingness to accept compensation to sell the same
good should be about equal (see Hanemann, 1991). But evidences over the past two decades
show that a significant gap exists between WTP and WTA. (Sinden and Knetsch, 1984;
Kahneman et al. 1990). One important explanation for the disparities between the WTP and
2
the WTA is labeled by Thaler (1980) “the endowment effect”. The endowment effect is the
increased value of a good to an individual when the good becomes part of individual’s
endowment. The person demands more to give up an object than she would be willing to pay
to acquire it. In order to study the effects of the endowment on the consistency of the choices
and on the direction of the preference reversals, we conducted our experiment in two groups:
the first group answered to the buying prices and the second group to the selling prices.
Cognitive Consistency Theory ( Lévy-Garboua, 1999; Blondel and Lévy-Garboua, 2005) ,
explains the disparity between WTA and WTP. According to this theory, uncertainty about
the exact value of an object for oneself can increase the divergence between WTP and WTA.
Unfamiliarity with a good can increase the uncertainty. In our experiment, we use two kinds
of good: organic and conventional products. We assume the fact that the organic products are
less present in the market compared to the conventional products, so they are less familiar for
the consumers.
The paper is organized as follows. Section 1 will present the context. The next section
will describe the experiment, and the last one presents the results.
In our experiments, instead of lotteries, we compare organic food and conventional food.
A wide majority of the consumers in France believe that the use of pesticides even in the
small amounts is dangerous for human health and the environment. The question of food
safety and organic farming is still an open one. It is not proved that organic farming preserves
the health while it preserves the environment. However, most of the people believe that
organic farming is the best strategy under food risks (Cowez, 2004). So we can say that in this
experiment, the “safer” option is the organic food and the “riskier” one is the conventional
food.
1. Context
Willingness to pays and experimental methods
To elicit the willingness to pay (WTP), we use an experimental method. The laboratory
experiment has some interesting advantages over other methods (survey, contingent
evaluation,…) because it involves real monetary payments and a situation wherein the
3
respondent chooses and subsequently consumes one or several food products. It creates a real
motivation for the participants to reveal their real preferences. Another advantage is that in a
lab environment we can compare the results of several preferences revealing procedures for
the same product.
There is a strong development of the experimental valuations of food quality and especially
food safety since the nineties (see the survey of this literature by Shogren, 2002, or Shogren et
al., 2002). These studies have used various goods and sources of risk. The goods were, for
instance, chicken (Shogren et al. 1999) or apples (Roosen et al. 1999). The sources of risk
were various food-borne pathogens (Salmonella, E. coli…) or the use or not of pesticides.
There are also some experimental studies that consider the food risk in a different way: low
stake and high probability risks instead of high stake and low probability risks. For example
the risk of having a though steak for dinner. ( Alfnes et al. 2005). There are many other goods
that these ones in the previous studies.
Some previous experimental results
The first experience has been conducted on March 2003 in ESA (Ecole suppérieure
d’agriculture, Angers) with 66 participants (Blondel and Javaheri, 2004). The objects were
bottles of wine and apples, with and without “ AB” label (agriculture biologique, i.e. organic
farming).
We asked the participants, first to give a WTP for each object, in several periods. Then we
presented the participants with a choice between two objects with a given price for each of
them. These given prices varied several times for each couple of product. The method used to
elicit the WTPs was the Becker, DeGroot and Marschak (1964) auction.
Looking at the WTPs we may be able to predict the choices that a rational subject will make.
For example when a subject who gave a WTP of 3 euros for organic apples and 2 euros for
conventional apples is asked to choose between conventional apple at 2.1 euros and organic
ones at 2.9 euros we can predict that she will choose organic apples. Hence a choice of
conventional apples is inconsistent with her WTP.
As expected, we found that the WTP for the organic bottles of wine and apples is significantly
greater than the conventional ones (mean of +25% for apples and +16.3% for bottles of wine).
We observed some inconsistent choices, but the rate of inconsistency, 13% is smaller than the
4
rate occurred in the lottery experiments, between 40% and 60% in the classical studies of
preference reversals.
In this experiment, all of the WTPs do not permit a test of consistency of the choice. In a
majority of the cases, both WTPs are lower than the prices of both products, proposed for the
choice questions. In that case no prediction of choices is possible. So 62% of the answers do
not permit tests, and the 13% of inconsistency represent only 5% among the remaining 38%.
A new experiment is conducted in March 2005. Several modifications are made related to the
previous experiment.
2. Experiment
A new experiment took place in 31 mars 2005, at the ESA laboratory for sensory analysis.
Sixty-nine subjects were recruited among the students of various disciplines in Angers
(France). They were endowed of 10 euros for their participation
We conducted the experiment in one day, with four sessions. Each subject was assigned to
participate in one session. Each session lasted approximately one hour. The subjects were
informed that they would participate in an individual decision making experiment, during
which they can buy some products. After an explanation about the rules of the experiment,
they were installed in the experiment lab that was partitioned to discourage communication
among participants. Thus each subject answered the questions on a computer.
The presented products were bags of conventional and organic apples (approximately one
kilo) and bottles of conventional and organic orange juices. All products could be found in the
local market.
They answered first to some general questions on their opinion about organic farming. The
session then proceeded with 4 pricing questions. Then they answered to some questions about
their age, budget and frequency of consumption before the other 20 decision questions about
the products.
The session 1 and 2 were dedicated to buying prices (36 participants) and the session 3 and 4
to selling prices (33 participants). Each participant answered to two kinds of decisions: prices
5
and choices. The participants to buying price sessions, were asked to indicate their WTP for
each good (organic and conventional apples, organic and conventional orange juices). Then
they answered to choice questions. A choice question consist on asking a participant whether
or not, she wants to buy a product in a given price. Figures 1 and 2 presents the framing of
both questions. The participants to selling price sessions answered to the pricing and choice
questions in the same order, but they indicated their WTA for each product. A choice question
for the selling price group, asks the participant to choose between an amount of money and a
product. In other words, there were two kinds of decisions: one where the subject has to
measure her preference (WTP and WTP), and one where the subject has to do a binary choice
(to buy or not, to sell or not), according with her preferences.
A computer program allows us to determinate the choice questions for each participant,
according to the answers she gave in the pricing questions. So the proposed prices (amounts
of money) depends on the buying (selling) price, which the individual indicated previously for
the same product. These two prices are equal to the price given by the subject herself less 0.1
euro in one hand, and more 0.1 in the other hand. Each answer to a pricing question
corresponds to two choice questions. Once, the offered price is higher than the participant’s
selling (buying) price, and once it is less. For example if a subject indicated a WTP of 1 euro
for a juice, in the choice phase, she will be presented with two questions. First, if she is ready
to buy the same juice at 1.1 euros and second, if she wants to buy it at 0.9 euro. The questions
of each phase were presented in a random order. (table 1).
[Insert Table 1 here]
We can expect that a rational subject refuses to buy (chooses the amount of money) for the
first choice question, and to buy (the product) for the second question. All of the questions
were repeated another time in order to check the robustness of the results. The order, in which
the questions were presented, is illustrated in figure 3.
[Insert figure 3 here]
When a subject had answered all the questions, her decisions were printed and she left the
Lab. Then, in another room, each participant draw a random number between one and twentyfour to determine which of the 24 decisions will be effective.
6

If the effective question was a choice one, the participant bought (sold) the product at the
given price, or not, according to her answer.

If it was a buying (selling) price question, the incentive mechanism used was the one
described by Becker et al. (1964). It consists in adding a random selling (buying) price,
here between 0.10 and 4 euros. If it was smaller than the WTP (greater than the WTA),
the consumer bought (sold) the product, at the random price. Otherwise, she didn’t buy
(kept) the product. This procedure can be shown to be incentive-compatible.
2. results:
General results
67% of 69 participants were female. The age of the subjects was between 18 and 45 years
with an average of 22 years. Table 2 shows some general information about the subjects’
characteristics. In order to compare the behavior of the two groups, it was necessary to check
that there was no significant difference between the socio-economic characteristics of the two
groups. The t-test results could not reject the hypothesis that demographics were equivalent
across group.
[Insert Table 2 here]
The opinion variable is an index calculated for each person on the base of her answers to the
opinion questions (all questions are presented in appendix 1). That varies between 16 and –16.
The more this index is high, the more the individual is worried about the pesticides.
Buying and selling prices
Figure 4 shows the average buying and selling prices for all products in two stages (all
questions were repeated in the second stage). We can see that for all products the average
price in second stage is lower than the average price for the first stage. The overall mean of
selling prices is 19% higher than the buying prices and this difference is significant ( t-test =
2.3872, P-value = 0.009)
[Insert Figure 4 here]
7
If we look at the WTA/ WTP ratio separately for organic and conventional products, we see
that the ratio of mean WTA and WTP for all organic products is 1,24 and 1,11 for
conventional products (t-test= , P-value= ). That confirms the idea that the gap is larger for
the unfamiliar products. (Table 5).
[Insert Table 5 here]
Consistency of the decisions
Now we look at the consistency of the choices in each group: buying price and selling price.
The average inconsistency rate is shown in the table 3. We can see that the average
inconsistency rate is 23% for the buying price group and 29% for the selling price group.
[Insert Table 3 here]
We can identify two types of inconsistencies for each group:

Buying price:
1) Refuse to buy a good at a price lower that the WTP (IC11).
2) Accept to buy a good at a price higher than WTP (IC12).

Selling price group:
1) Choose the money, between a good, and a price lower that one’s WTA (IC21)
2) Choose a good, between a good and an amount of money higher than one’s WTA (IC22)
These results are shown in table4. (In the first group the rate of CI11 (14%) is greater than
IC12 (9%). It means that) Among the inconsistent participants in the first group, those who
chose not to buy a good at a price p lower than their maximum buying price (14%) , are more
than the participants who chose to buy a good at a price higher that their maximum buying
price (9%). But among the selling group, the rates of inconsistencies in both directions are
equal (15%)
[Insert Table 4 here]
Figure 5 summarizes the average rates of inconsistencies.
[Insert figure 5 here]
8
Finally, we done a regression to see which factors influence the inconsistency. The dependent
variable is the rate of inconsistency, and the independent variables are age, gender, income,
consumption of the product, mean WTP (or WTA), and a new variable: stability. This last one
is defined as follows:

1 if the subject does not give the same WTP (WTA)

0 otherwise.
We done 8 regressions, for each of the four products (apples, orange juices; conventional and
organic farming) and both groups. The main conclusion is that the significant factor is the
stability. Hence, we have redone the figure 5 with two cases: stability or not. The results are
in figure 6.
[Insert figure 6 here]
The rate of inconsistency is significantly greater when the WTP or WTA are unstable. Hence,
the preference reversal could be due in part to the instability of the preferences.
We will cut in two parts both groups, in order to compare the WTA/WTP disparity. The
hypothesis that we will test is he following: the WTA/WTP disparity decreases with the
knowledge of the product. Indeed, we observe that this disparity is significantly greater for
organic farming products. In both groups, we can separate the less experienced subjects from
others.
Develop a model of decision with cognitive consistency theory (Lévy-Garboua, 1999;
Blondel and Lévy-Garboua, 2005). This theory explains the disparity between WTA and
WTP.
Each product seems at first sight a sure option but in reality, the level of quality expected by a
subject depends on the knowledge of this product by the subject. Hence, the utility of the
product depends of a risky quality. We will assume that this quality is a random variable
uniformly distributed on an interval
q, q
9
*****************
References
Alfnes F., Rickertsen K., Ueland O. (2005), “ experimental evidence of risk aversion in
consumer markets: the case of beef tenderness”, paper presented at the XIIth International
Congress of European Association of Agricultural Economists. Copenhagen, Denmark,
August 24-27, 2005
Becker G.M., De Groot M.H. and Marschak J. 1964. Measuring utility by a single-response
sequential method. Behavioral Science. 9: 226-232.
Blondel S. and M. Javaheri. 2004. Valuing Organic Farming: an Experimental Study of the
Consumer. Acta Horticulturae. 655: 245-252.
Casey J.T. (1991). “Reversal of the Preference Reversal Phenomenon”, Organizational
Behavior and Human Decision Processes 48, 224-251.
Casey J.T. (1994). “Buyers’ Pricing Behavior for Risky Alternatives: Encoding Processes and
Preference Reversals”, Management Science 40, 730-749.
Cowez A. 2004. Organic farming and food safety, Student thesis, Institut national
d’horticulture.
Hanemann, W. M. (1991), "Willingness to pay and willingness to accept: How much can they
differ?", American Economic Review 81:635-647.
Kahneman, D., J. Knetsch and R. Thaler (1990), "Experimental tests of the endowment effect
and the Coase theorem", Journal of Political Economy 98: 1325-1348.
Knetsch, J. and J. A. Sinden (1984), "Willingness to pay and compensation demanded:
Experimental evidence of an unexpected disparity in measures of values", Quarterly
Journal of Economics 99:507-521.
Lichtenstein S., and P. Slovic 1971. Reversals of preferences between bids and choices in
gambling decisions, J. of Experimental Psychology. 89: 46-55.
Lindman H. 1971. Inconsistent preferences among gambles, Journal of Experimental
Psychology. 89: 390-397.
Roosen J., Fox J., Hennessy D. and A. Schreiber. 1998. Consumer’s Valuation of Insecticide
Use Restrictions: An Application to Apples. Journal of Agricultural and Resource
Economics. 23: 367-384.
10
Shogren J., Fox J., Hayes D., and J. Roosen. 1999. Observing Choices for Food Safety in
Retail, Survey, and Auction Markets. American Journal of Agricultural Economics. 81:
1192-1199.
Shogren J. 2002. Experimental Methods and Valuation, Handbook of Environmental
Economics (Mäler & Vincent, eds.) Elsevier.
Shogren J., Hayes D.H., Fox J and T. Cherry. 2002, What am I bid for food safety? Lessons
from a Decade in the Lab, Choices, 16-21.
Slovic P. 2000. The Construction of Preference, in Choices, Values, and Frames, ed. by
Kahneman D. and Tversky A. Russell Sage Foundation. Cambridge University Press, 489502.
Thaler, Richard (1980) “Toward a Positive Theory of Consumer Choice.” J. Econ. Behavior
and organization 1 (March 1980): 39 – 60
11
Tables
Table 1. How the preferences are elicited
Buying price (WTP)
WTP + 0.1
WTP – 0.1
Pricing question:
Choice question 1:
Choice question 2:
A juice at 1.1, do you buy it?
A juice at 0.90 do you buy it?
Selling price (WTA)
WTA + 0.1
WTA – 0.1
Pricing question:
Choice question 1:
Choice question 2:
Selling price for a juice?
Choose between 1.1euro and a Choose between 0.9 euro and
Answer = 1 euro
juice
Buying price for a juice?
Answer = 1 euro
a juice
Table 2. Summery statistics of characteristics
variable
gender
definition
1 = female,
Group1
Group2
t-test
P-value
(mean)
(mean)
0.67
0.67
0
1
0= male
age
Age in years
21.27
23.11
-2.1803
0.0327
income
Household monthly income level
2.27
2.55
-0.629
0.5317
1= less that 400 €
2 = 400 to 600€
3 = 600 to 800€
4 = 800 to 1000€
5 = 1000 to 1200€
6 = 1200 to 1400€
7 = more than 1400€
12
Opinion
Subjects
Number of participants
3.48
2.44
33
36
0.7024
0.4849
Table 3. the average inconsistency rate.
buying
selling
CN
OF
CN juice OF juice CN
apple
apple
First
23%
27%
26%
second
23%
18%
17%
average
23%
OF
CN juice OF juice
apple
apple
29%
40%
28%
26%
28%
20%
29%
36%
22%
29%
29%
Table 4. Direction of inconsistencies
Buying price
Selling price
Refuse to buy at
Buy at
P<WTP
P>WTP
number
74
46
percentage
14%
9%
total
Choose P<WTA
Choose the good
over the good
over P>WTA
120
87
85
172
23%
15%
15%
30%
Table 5: WTA/WTP ratios for each product.
WTP
WTA
WTA/WTP
OF apple
1,555
1,95
OF juice
1,365
1,675
CN apple
1,175
1,37639
CN juice
1,01
1,06
5,105
6,06139
Mean
1,254 Mean for OF
1,227
1,240
1,171 Mean for CN
1,049
13
1,110
total
Figures
Figure 1. Example of a pricing question
PRICE:
A BAG OF CONVENTIONAL APPLES
• What is the maximum price you are willing to pay for a
bag of conventional apple?
 your price =
euros.
• A random Buying price included between 0.1 € et 4 €
will be drown:
• if:
– your price  selling price: you buy at selling price
– your price  selling price : you can not buy
Figure 2. Example of a choice question:
CHOICE :
A BAG OF CONVENTIOANL APPLES
You must decide, to buy or not, a bag of conventional
apples, at the proposed price:
price of a bag of conventional apples = 1.1 euros
buy
not to buy
click on the correspondent case.
14
Figure 3. How proceed the experiment
6 opinion questions
buying
selling
Pricing
Stage 1
(4 questions)
WTP
WTA
General questions
Choice (8
questions)
Pricing
Choose between :
• nothing
• buy at WTP
+ or - 0, 1
Choose between :
•product
• WTA + or - 0, 1
WTP
WTA
(4 questions)
Stage 2
Choice (8
questions)
At the end
Chose between :
• nothing
• buy at WTP
+ or - 0, 1
Chose between :
•product
• WTA + or - 0, 1
Random selection
of one question
between 24
If pricing =>
BDM procedure
If choice : buy if chosen
If pricing =>
BDM procedure
If choice : receive the chosen
15
Figure 4. Main results
average bying and selling prices
2,5
2
Buying First
Buying Second
Selling First
Selling Second
1,5
1
0,5
0
AB
apple
Cl
apple
AB
juice
CL juice
Figure 5. Rates of inconsistencies
Inconsistent:
120 (22.7%)
WTP group:
528
Consistent :
408
Total choice
answers:
1104
Inconsistent:
172 (29.6%)
WTA group:
576
consistent:
404
16
Figure 6. Rates of inconsistencies
WTP stable:
200
WTP group:
528
Total choice
answers:
1104
WTP unstable:
328
WTA stable:
248
Inconsistent:
31 (15%)
Consistent :
169 (85%)
Inconsistent:
89 (27%)
Consistent :
239 (73%)
Inconsistent:
59 (24%)
Consistent :
189 (76%)
WTA group:
576
WTA unstable:
328
Inconsistent:
113 (34%)
Consistent :
215 (66%)
17
Appendix 1
Q1- d’après ce que vous savez, les pesticides tels qu’utilisés aujourd’hui dans les cultures,
représentent-ils un danger très important, assez important, peu important ou pas important du tout
pour:
Très important/ assez
Pas important/ pas important
Sans réponse
important
du tout
L’eau
94%
6%
0%
Sol
86%
14%
0%
Air
45%
49%
6%
Animaux
56%
43%
2%
Homme
63%
38%
0%
Q2- Avec la quelle de ces deux opinions vous êtes le plus d’accord ?
Les pesticides présentent peu de
Les pesticides présentent des
risque pour l’homme et
risques importants pour
l’environnement si on les utilise
l’environnement et pour l’homme,
selon les doses prescrites
quelles que soient les doses
Sans réponse
d’utilisation
57%
38%
6%
Q3- Face à des risques sur l ’environnement et pour l ’homme, quelle est la solution qui vous paraît la
plus efficace?
Augmenter de façon très
Continuer à traiter les cultures
importante la part de l’agriculture
mais avec des quantités de
biologique, même si les coûts de
pesticides moins importantes
production sont plus élevés
18
Sans réponse
41%
55%
4%
Q4-Selon vous, y a-t-il des résidus de pesticides sur les fruits et légumes que l ’on trouve le plus
couramment dans le commerce?
Oui, c ’est certain
oui, sans doute
Non, sans doute pas C ’est certain que
sans opinion
non
49%
36%
14%
0%
19
0%
Q5- S ’il y a des résidus sur ces produits, pensez-vous qu’en général :
Ils sont en quantité susceptible
ils sont en si petite quantité qu’ils
d ’être nuisible pour la santé
n ’ont aucun effet sur la santé
41%
54%
sans opinion
6%
Q6- Si des scientifiques indépendants vous disent qu’en dessous d ’une certains dose, des résidus de
pesticides dans les aliments n ’ont aucune incidence sur la santé, avez-vous tendance:
A les croire
A ne pas les croire
sans opinion
74%
19%
7%
20