Professional Documents
Culture Documents
Food Policy: J. Roosen, A. Bieberstein, S. Blanchemanche, E. Goddard, S. Marette, F. Vandermoere
Food Policy: J. Roosen, A. Bieberstein, S. Blanchemanche, E. Goddard, S. Marette, F. Vandermoere
Food Policy
journal homepage: www.elsevier.com/locate/foodpol
a r t i c l e
i n f o
Article history:
Available online 20 January 2015
Keywords:
Evaluation
Nanotechnology
One-and-a-half-bound dichotomous choice
Self-protection
Trust
Willingness to pay
a b s t r a c t
We analyze the role of trust in the evaluation of a new food technology, namely nanotechnology. A literature review in the social and economic sciences reveals that many different trust concepts are available.
The economics literature suggests that trust can lead to lower efforts of self-protecting behavior. Translating this concept into the framework of willingness to pay (WTP) for food products allows for the derivation of hypotheses on the workings of trust. We show that WTP for new food characteristics increases
with trust also when new information about the technology is revealed. The results are conrmed with
online survey data for Canada and Germany and experimental data in Germany.
2014 Elsevier Ltd. All rights reserved.
Introduction
The development of integrated, long supply chains and some
food technologies has made modern food businesses more vulnerable to safety incidents and consumer trust in the food industry
has been challenged by a series of scandals. Consumers have
become skeptical of food innovations and industrialization. Even
when responding to the growing demand for convenience and
safety, the successful introduction of new food technologies has
become a major challenge.
It has been shown that trust is a construct that helps people to
accept risks in the face of moral hazard. Hence, trust contributes to
economic progress (Arrow, 1974). However, despite this recognized importance of trust in the economics literature, most original
contributions and those that operationalize trust as a measurement construct have come from the eld of other social sciences,
such as sociology (e.g., Luhmann, 1968/2000; Giddens, 1990). Here
trust is a source of social capital that helps to reduce complexity
and to facilitate interaction. Giddens (1990), for example, posits
that trust in expert systems is a mechanism to reduce complexity:
when a laypersons knowledge is inadequate, the person will retain
his or her ontological security by trusting experts. In a context of
food safety crises and the development of new food technologies,
trust is considered to be a key concept (Berg, 2004; de Jonge
Corresponding author at: TUM School of Management, Marketing and Consumer Research, Alte Akademie 16, 85350 Freising, Germany. Tel.: +49 8161 71
3318; fax: +49 8161 71 4501.
E-mail address: jroosen@tum.de (J. Roosen).
http://dx.doi.org/10.1016/j.foodpol.2014.12.004
0306-9192/ 2014 Elsevier Ltd. All rights reserved.
et al., 2008; Frewer et al., 1996; Kjrnes et al., 2007; Renn and
Rohrmann, 2000; Sassatelli and Scott, 2001).
The objectives of this paper are twofold. From extant literature,
denitions of trust are reviewed with regard to food technologies
and an attempt is made to relate trust to consumers willingness
to pay (WTP) for altered food characteristics. Secondly, the
relationship between trust and WTP is assessed for a new food
attribute that is introduced using nanotechnology. We analyze
whether trust correlates with the acceptance of a functional food
attribute (here: vitamin enrichment and protection) and if this
evaluation changes when consumers learn that the attribute has
been created by means of nanotechnology.
Food nanotechnology is regulated under existing legislation
(European Commission, 2012; Government of Canada, 2013) and
a case by case approach is applied through a pre-market approval
system (European Commission, 2012). A review of approaches to
the regulatory governance of nanotechnology up to 2009 can be
found in Pelley and Saner (2009). However, the application of
nanotechnology in the food domain is surrounded by high levels
of scientic uncertainty with several studies pointing to possible
negative long term effects (Wang et al., 2006; Oberdrster et al.,
2005). The application of nanotechnology in the food industry is
still limited and new and rather unknown to consumers. In this
context of low knowledge, high complexity and high uncertainty
nobody retains the authority of better knowledge (Luhmann,
1993) and the safety of the food market increasingly depends on
the decisions of the responsible actors (Fischler, 1988). Consumer
acceptance of such new, complex technologies is likely to depend
on how much they trust these actors.
76
77
Nanotechnology refers to materials, systems and processes which exist or operate in the range of about 1
100 nm (nm). One nanometer (nm) is one millionth of a
millimeter (mm). Materials at this scale show novel properties that lead to novel applications in diverse fields such
as medicine, cosmetics, biotechnology, energy production and environmental science. There is uncertainty
regarding how nanomaterials may interact with human
health and the environment.
Nanotechnology offers new opportunities for food industry application. Manufactured nanomaterials are already
used in some food products, nutritional supplements,
and food packaging applications. Nanotechnology allows
for the improvement of barrier functions in food packaging to reduce UV-light exposure or microbial growth and
thus extend the shelf-life of many food-products. Furthermore, nano-biosensors are able to control the foods level
of freshness by indicating spoiled food to the consumers
by means of color change. There is not much known about
the effects on human health and environment.
78
Willingness to pay I
In what follows we will present you information about two pure orange juices sold in one litre
bottles. On the market, the average price of this type of orange juice varies between $1.75 and
$2.50 per litre.
Orange juice A
Orange juice B
Willingness to pay II
Nanotechnology offers new opportunities for food industry application. Manufactured
nanomaterials are already used in some food products, nutritional supplements, and food
packaging applications.
Two examples in development are the two orange juices than have already been presented to
you above. (In the market, the average price of this type of orange juice varies between $1.75
and $2.50 per litre)
Orange juice A
Orange juice B
for their willingness to buy the juices at the varying prices with
yes, no or maybe. Only afrmative yes-responses were counted
in determining WTP.
(1) The respondent accepts the product neither at the low nor at
the high price. The probability for the observation is
pnm GPLi ; h.
(2) The respondent accepts the product at the low price but not
at the high price. The probability is pym GP Ui ; h GPLi ; h.
(3) The respondent accepts the product at the high price. The
probability is pyy 1 GPUi ; h.
In the choice question, a maybe option was included: these
maybe responses have been counted as no answers in the
econometric estimation.
The resulting log-likelihood function for the sample i = 1, . . . , N
is
Analysis of WTP
ln Lh
For the analysis of the WTP variables in the OOHB format, we
follow Cooper et al. (2002). Let the true, unknown WTP of consumer i be denoted by Ci. Ci can be cast in the random utility framework and is a function of personal characteristics of respondent i.
I.e., if WTP follows a cumulative distribution function G(Ci, h), then
the parameter vector h describes the parameters of the distribution, such as the mean and variance.
Two different prices are proposed to respondents: PU(=$2.5/
1.80) and PL (=$1.75/0.90). There are three possibilities:
N h
X
nn
yn
di lnG PLi ; h di ln G PUi ; h G PLi ; h
i1
i
yy
di ln 1 G PUi ; h
The logistic function is commonly used in dichotomous choice
estimations due to its close resemblance to the normal distribution
function and its computational ease. We use a logistic function for
G(; h) and assume that the mean is a linear function of K personal
characteristics of the respondent and the parameter vector to be
estimated. Hence, expected WTP above the price can be derived as
79
EC i a0 b1 X 1i . . . bK X Ki cPi
with a0 as a constant and the parameter c to price measuring the
marginal utility of income. Average WTP for the sample average
can be estimated by estimating E[Ci] = a0 cPi (without explanatory variables), so that WTP = a0/c. The standard error of the WTP
estimate is obtained by the delta method.
Regarding the experimental WTP data, we adopt a standard
Tobit model on the pooled observations. WTPi is used as the dependent variable as censored at PU = 1.80 and PL = 0.90, so that
8
if 0:90 6 C i 6 1:80
>
< Ci
WTPi 0:90 if C i < 0:90
>
:
1:80 if C i P 1:80
Table 2
Descriptive statistics Germany online survey and lab experiment.
Variable
Websurvey (N = 750)
Mean
Male
Income
Age
0.51
1995.43
45.38
4.12
3.74
3.72
2.87
0.48
Experiment (N = 143)
Std. dev.
Mean
Std. dev.
1265.59
15.36
0.45
2 577.77
45.74
1 630.76
14.35
0.86
0.85
1.04
0.63
3.80
3.99
3.92
2.90
0.48
0.78
1.01
0.73
0.62
respondents. Compared to the population of the city the experimental sample is somewhat older and has a higher income.
Results
An online survey was conducted with 615 English speakers
residing in Canada. Table 1 provides descriptive statistics on several sociodemographics of the sample and general trust measures.
About 50% of the sample are male, the average income is Ca$70 578 and the mean age of the sample is 50.1 years. Compared
to the Canadian population survey respondents are somewhat
older and better off (Matin et al., 2012).
Among the food choice motives according to Steptoe et al.
(1995) we consider importance of price (M_price), importance
of nutrient content (M_nutrition) and importance of naturalness (M_natural) as relevant. Items are evaluated on a ve-point
Likert scale. We report summated scores with M_price containing
two, M_nutrition six, and M_natural three items. The nutrient and
the naturalness motive are with average values of 3.45 and 3.39 of
about equal importance. The price motive is slightly more important (3.69). The average institutional trust in food has a mean value
of 2.75 (on a 5-point scale) and 55% of the sample show low trust
(point value below three).
The German online survey was conducted with a sample of 750
respondents between February and April 2009. Table 2 provides
descriptive statistics. The average age is 45 years, 51% of the sample are male. Net monthly income is on average 1 995. Compared
to the German population survey respondents are less well-off
(Federal Statistical Ofce of Germany, 2013). The price motive
turns to be most important (4.12) compared to nutrition (3.74)
and naturalness (3.72). The mean trust score is 2.87, yielding a
low level of trust for 48% of the sample.
Table 2 also provides some summary statistics for the experiment conducted in Munich, Germany. The lab sessions took place
between January and February 2009 with 143 participants. The
majority of participants (55%) are female and the average monthly
net income is above that of the online survey sample. The
experimental participants are of similar age as the online survey
Table 1
Descriptive statistics Canada online survey.
Variablea
Denition
Mean
Std. Dev.
Male (N = 613)
Income (N = 610)
Age (N = 615)
1 = Male, 0 = Female
Total household income in Ca-$ (annual)
In years
0.49
70577.87
50.10
33937.12
13.79
3.69
3.45
3.39
2.75
0.91
0.85
1.11
0.76
Because of varying number of missing data, the number of observation is indicated in parentheses (N=).
0.55
80
Table 3
Risk perception in Canada and Germany, online survey.
Canada
Germany
Total
Mean(Std.
Dev.)
Low Trust
Mean(Std.
Dev.)
High Trust
Mean(Std.
Dev.)
Total
Mean(Std.
Dev.)
Low Trust
Mean(Std.
Dev.)
High Trust
Mean(Std.
Dev.)
2.21(0.84)
2.17(0.84)
3.84(1.02)
2.25(0.84)
2.18(0.83)
3.99(1.00)
2.14(0.84)
2.14(0.85)
3.65(1.02)***
2.73(1.04)
2.53(0.97)
4.41(0.92)
2.86(1.08)
2.63(1.02)
4.50(0.93)
2.61(0.98)***
2.44(0.91)**
4.33(0.91)***
3.53(1.15)
3.52(1.15)
3.54(1.00)
3.83(0.99)
3.13(1.13)
3.74(1.11)
3.63(1.11)
3.79(0.96)
4.04(0.92)
3.44(1.12)
3.27(1.15)***
3.37(1.17)**
3.23(0.98)***
3.57(1.02)***
2.82(1.05)***
3.95(1.12)
4.29(0.97)
3.62(0.97)
4.44(0.88)
3.66(1.13)
4.03(1.16)
4.35(0.99)
3.75(1.04)
4.56(0.83)
3.82(1.13)
3.87(1.08)***
4.24(0.95)**
3.49(0.95)***
4.34(0.90)***
3.49(1.11)***
Refer to the difference between high and low samples being signicant at the 10% level as tested by a MannWhitney U-test per country.
Refer to the difference between high and low samples being signicant at the 5% level as tested by a MannWhitney U-test per country.
Refer to the difference between high and low samples being signicant at the 1% level as tested by a MannWhitney U-test per country.
Table 4
Parameters of logistic one-and-a half bound estimation, Canada (N = 510).
Juice A (Vitamin D enriched)
Without Nano
With Nano
Without Nano
With Nano
Constant
Trust
Gender (male = 1)
M_price
M_nutr
M_nat
Price
2.747*** (0.586)
0.331*** (0.116)
0.068 (0.168)
0.253** (0.106)
0.500*** (0.140)
0.030 (0.107)
2.265*** (0.149)
1.010 (0.580)
0.387*** (0.115)
0.124 (0.173)
0.020 (0.111)
0.326** (0.145)
0.107 (0.113)
1.706*** (0.136)
1.777*** (0.578)
0.313*** (0.116)
0.198 (0.169)
0.081 (0.105)
0.342** (0.141)
0.032 (0.106)
1.996*** (0.147)
0.443 (0.614)
0.364*** (0.128)
0.339* (0.187)
0.028 (0.121)
0.245 (0.155)
0.169 (0.122)
1.512*** (0.145)
LogL
527.337
502.689
514.423
442.404
Table 5
Mean WTP and condence intervals, Canada (N = 510).
WTP without
Nano
WTP with Nano
Juice A (Vitamin D
enriched)
Juice B (Vitamin C
protected)
product B may be perceived as less benecial given the easily alternative sources of vitamin C. This coincides with results by Zhou
(2013) that show a lower WTP for nanopackaging in a hypothetical
choice experiment conducted on US consumers compared to a
functional health benet offered by nanodrops. The lower preference for nanopackaging may also explain why WTP for juice B
decreases by a larger amount as compared to juice A when respondents learn that the juice is produced by means of nanotechnology.
Indeed, the decrease in mean WTP is much larger for juice B
($0.428) than juice A ($0.245).
Tables 6 and 7 show the results for the German online survey.
The sample size reduces to 740 respondents due to missing data.
Results are fairly similar compared with Canada. Male compared
to female respondents have a higher WTP (not signicant for Juice
A before Nano information). Price is a food choice motive that
reduces WTP for Juice B without Nano information whereas the
motive nutrition (M_nutr) increases it for Juice A. Finally trust
81
Without Nano
with Nano
Without Nano
With Nano
Constant
Trust
Gender (male = 1)
M_price
M_nutr
M_nat
Price
0.406 (0.536)
0.418*** (0.118)
0.182 (0.147)
0.056 (0.094)
0.204* (0.121)
0.010 (0.098)
2.583*** (0.150)
0.836 (0.638)
0.709*** (0.131)
0.603*** (0.171)
0.084 (0.107)
0.138 (0.138)
0.147 (0.108)
2.315*** (0.179)
1.334** (0.567)
0.363*** (0.122)
0.369** (0.152)
0.232** (0.099)
0.033 (0.121)
0.152 (0.099)
2.791*** (0.189)
0.287 (0.579)
0.672*** (0.127)
0.579*** (0.163)
0.156 (0.104)
0.127 (0.132)
0.118 (0.103)
2.276*** (0.161)
LogL
663.173
525.655
596.815
576.064
Table 7
Mean WTP and condence intervals Germany (N = 740).
WTP without
Nano
WTP with Nano
Juice A (Vitamin D
enriched)
Juice B (Vitamin C
protected)
Because WTP has been elicited using a price list, empirical distributions of WTP can be described. The WTP data is described in
Table 8. Average WTP is 1.154 for Juice A and 1.004 for Juice B
(including zero bids). WTP decreases with information about the
use of nanotechnology. Hence also in the experiment, the use of
the technology is evaluated as bad news. Juice B is rejected by
more consumers, however, the number of consumers decreasing
their bid to zero is not as large (9) as in the case of Juice A (41).
Splitting the sample into consumers with low and high trust, it
can be observed that the low trust consumers are more likely to
decrease their bid to zero when they learn about the application
of nanotechnology, even if only slightly. E.g. for juice A the number
of nonengaged consumers increases from 11 to 33 in the sample of
low trust respondents and from 7 to 26 in the sample of high trust
respondents.
Fig. 2 shows the empirical cumulative distribution functions
(cdf) of WTP before and after information about the use of nanotechnology for juice A and B. Values below the lower bound of
the price list at 0.90 are coded as a WTP of zero, whereas yesresponses to a price of 1.80 are recorded as 1.80. To explain
the gure, we carefully discuss the graph for Juice A and the low
trust sample (upper left). As shown by the red, solid line, about
12.5% of the sample has a WTP below 0.90. As the WTP value
increases, the share of respondents with a WTP of a given value
and below increases stepwise, as we asked for WTP in 10 cents
intervals. The blue, dotted line shows the cdf after the information
that Juice A was manufactured by means of nanotechnology. The
upward and left shift of the step functions indicates the increase
in the proportion of respondents denoting a zero WTP (now 50%
of the sample). Indeed the maximum WTP observed is 1.60, while
it was 1.80 before the information.
In all cases the WTP distribution shifts leftwards after information about the use of nanotechnology. Comparing the low trust and
Table 8
Description of WTP data, experiment, Germany.
Juice A (Vitamin D enriched)
Without Nano
With Nano
Without Nano
With Nano
Average WTP
incl. zero bidsa
excl. zero bids
1.154
1.298
0.756
1.280
1.004
1.290
0.878
1.283
18/119
11/54
7/65
59/80
33/32
26/48
36/101
22/42
14/59
45/93
26/38
19/55
Bids with no responses at all prices have been coded as 0.00. Bids with yes responses at all prices are coded as 1.80.
82
0.8
0.6
CDF
0.4
0.2
0
0.5
1.5
WTP
Aer Informaon A
before Informaon A
Aer Informaon A
1.5
Before Informaon A
0.8
0.6
0.6
CDF
0.4
0.2
0
WTP
0.8
CDF
0.5
0.4
0.2
0.5
1.5
WTP
Aer Informaon B
before Informaon B
Aer Informaon B
0.5
WTP
1.5
Before Informaon B
high trust samples for juice A and juice B, it becomes obvious that
before information (solid lines) the distribution of WTP for the high
trust sample is to the right of the one for the low trust sample. In
all four graphs it is consistently observed that the information
about the use of nanotechnology leads to a leftward shift in the
cumulative distribution function. Based on this descriptive result
it appears that the distribution of WTP after information is
rst-degree stochastically dominated by that from before the information. This means that information about the use of a new technology leads to a welfare decrease for consumers. As before, the
information is considered bad news. Secondly, we observe that
high levels of trust not only increased (rightward shift) the initial
WTP distribution of respondents but also reduced the size of the
leftward shift in the distribution due to new information. Hence,
trust seems to yield a double dividend, in the form of a higher original acceptance of nanotechnology and less reduction after additional information is provided about the technology.
Table 9
Parameters of Tobit estimation, experiment, Germany.
Juice A (Vitamin D
enriched)
N = 268
Juice B (Vitamin C
protected)
N = 267
Constant
Information
Trust
Gender
(male = 1)
M_price
M_nutr
M_nat
Sigma
1.231*** (0.199)
0.206*** (0.042)
0.069** (0.033)
0.020 (0.043)
1.601*** (0.216)
0.050 (0.045)
0.061* (0.037)
0.028 (0.046)
0.039 (0.030)
0.070** (0.033)
0.085*** (0.026)
0.322*** (0.018)
0.112*** (0.032)
0.025 (0.037)
0.030 (0.028)
0.350*** (0.021)
LogL
132.934
152.556
Conclusion
83