-
Submitted Article
Do Consumer Responses to Media Food SafetyInformation Last?
Robin Dillaway, Kent D. Messer*, John C. Bernard, andHarry M.
Kaiser
Robin Dillaway is a graduate of the Department of Food and
Resource Economicsat the University of Delaware. Kent D. Messer is
an associate professor andJohn C. Bernard is a professor in the
Department of Food and ResourceEconomics, and the Department of
Economics, respectively, at the University ofDelaware. Harry M.
Kaiser is the Gellert Family Professor of Applied Economicsand
Management at Cornell University.
*Correspondence may be sent to: [email protected].
Submitted 18 June 2010; accepted 17 March 2011.
Abstract Using experimental methods with adult subjects from the
mid-Atlantic region of the United States, this research examines
both the short- andlonger-term impacts of media information on
consumer purchasing behavior.Subjects in the treatment group were
given food safety information about poultryfrom a popular consumer
magazine. Willingness to pay (WTP) estimates werethen elicited for
two types of chicken breasts: (1) a leading-brand that was
identi-fied in the information treatment as having a high incidence
of Campylobacterand Salmonella bacteria; and (2) a lesser known
brand, which was reported asbeing relatively free of harmful
bacteria. Results indicated that both negative andpositive food
safety information significantly impacted consumers’ WTP for
saferchicken compared to the reportedly less-safe leading-brand
chicken. These changesin behavior persisted throughout the
seven-week study period.
Key words: Consumer behavior, food safety, experimental
economics,media information.
JEL Codes: Q13, D83, C91.
Introduction
The Centers for Disease Control and Prevention (CDC) estimates
thatthere are approximately 76 million illnesses, 325,000
hospitalizations, and5,000 deaths annually caused by food-borne
diseases in the United States(Mead et al. 1999). The U.S.
Department of Agriculture’s EconomicResearch Service estimates that
food-borne illnesses from the top fivepathogens affecting humans
cost society $6.9 billion annually (Crutchfieldand Roberts, 2000).
Moreover, a recent study estimates that food-borneillness has a
societal cost of $357 billion annually (Roberts, 2007).
Clearly,
# The Author(s) 2011. Published by Oxford University Press, on
behalf of Agricultural and AppliedEconomics Association. All rights
reserved. For permissions, please
email:[email protected].
Applied Economic Perspectives and Policy (2011) volume 33,
number 3, pp. 363–383.doi:10.1093/aepp/ppr019
363
-
food-borne illnesses are a health risk as well as an economic
burden onsociety, yet the effect of food-borne illnesses on
consumer behavior is notwell understood. The issue of whether and
how quickly consumers returnto their previous behavior after
receiving media food safety information isimportant to consider
when studying the effect of information on consum-ers’ purchasing
behaviors. However, research on how quickly the impactof food
safety information decays is scarce. Accordingly, the focus of
thisresearch is to provide a better understanding of consumer
reactions tomedia food safety information over time.
Media reports about food safety concerns are relatively
common.Consider for example, the 2006 Escherichia coli outbreak in
spinach, whichresulted in 204 illnesses, 104 hospitalizations, and
3 deaths (Calvin 2007).In September 2006, the Food and Drug
Administration (FDA) advised con-sumers not to eat bagged spinach
and expanded the warning the follow-ing day to fresh spinach. This
resulted in a five-day period during whichno fresh spinach was sold
in the United States, while California spinachremained off the
market for ten days (Calvin 2007). In the longer term,sales of
bagged spinach were depressed for months afterward. Comparedto the
same period from the previous year, retail sales of bagged
spinachwere down 27% five months after the outbreak (Calvin
2007).
Knowing more about the effects of food safety information on
consumerbehavior is of particular importance to government
agencies. Agenciessuch as the CDC and FDA issue numerous warnings
about specific foodproducts. For instance, the CDC reports that in
2006, there were 1,247food-borne outbreaks (CDC 2008). Given the
large number of reportedoutbreaks, their impact on consumers can be
important.
Media reports are an important source of information about food
safetyissues for consumers. The degree of coverage a food safety
incidentreceives is likely to influence consumer decisions.
However, little isknown about how consumers react to situations
where there is an initialburst of media information about the
safety of a food product and thenrelatively little information is
provided as the media focuses its attentionon other topics, which
is common practice, as food safety tends to receivesporadic
coverage in most media outlets.
This study differs from most of the previous literature in that
it seeks tounderstand both short- and longer-term consumer
responses to food safetyconcerns by eliciting willingness to pay
(WTP) for poultry products usingexperimental economics methods.
This study expands upon previousexperiments on food safety with its
multiple sessions, which allows for anexamination of the dynamic
processes that exist in the real world. Mostprevious experiments,
(for example, Hayes et al. 1995; Lusk and Schroeder2002; and
Thomsen and McKenzie 2001,) which consist of a single,
isolatedsession, are unable to show consumers’ changing attitudes
and WTP overtime. To our knowledge, the seven-week study period
used here is thelongest used by an experimental study of this kind.
Other studies (forexample Saghaian 2007; Beach et al. 2008) of
consumer response to foodsafety scares have traditionally relied
upon aggregate market data and donot track specific individuals to
see how their behavior changes over time,as the experimental
framework here allows.
Our study involved 110 adult participants who were randomly
assignedto either the control group or the treatment group. Each
subject partici-pated in three sessions over seven weeks.
Participants in the control
Applied Economic Perspectives and Policy
364
-
treatment were given no food safety information. Participants in
the foodsafety treatment were given information from a Consumer
Reports maga-zine article, which stated that during the testing of
leading brands ofchicken (identified specifically by Consumer
Reports as Perdue, Tyson,Foster Farms, and Pilgrim’s Pride)1 high
rates of contamination of harmfulbacteria were found. Ranger brand
chicken, however, was relatively cleanof these bacteria (Consumer
Reports 2007). Analysis of the WTP data indi-cated that
participants in the treatment group were willing to pay morefor the
Ranger chicken (a relatively safer product). Results from this
studyindicated that consumers were strongly affected by food safety
informa-tion both immediately and over time.
Relevant Literature
The impact of food safety information on consumers’ WTP has
beenwell established in both experimental and empirical studies. In
general,the experimental studies have used controls available in a
laboratory tomeasure consumers’ WTP at a single moment in time,
while the empiricalstudies have employed econometric techniques to
examine the change inconsumer behavior over time. That food safety
information affects con-sumer purchasing decisions is well
established in the literature with somestudies focusing on the
effect of general information alone, others focusingon positive
information, others on negative information, and some using
acombination of both. The experiments of Hayes et al. (1995)
demonstratedthat the availability of information can change
consumers’ purchasinghabits in response to perceived risk, as
consumers in this study werewilling to pay a premium for safer
food. The authors also found thatexperimental subjects tended to
underestimate the probability of a food-borne illness.
Several empirical studies have also shown food demand to be
affectedby food safety concerns. For instance, Piggitt and Marsh
(2004) found arelatively small downward demand shift in response to
food safety con-cerns, while Marsh, Schroeder, and Mintert (2004)
found a significantdownward demand shift in response to Food Safety
Inspection Servicerecalls. Literature on food product recalls has
also shown negative marketimpacts (for example, Lusk and Schroeder
2002; Thomsen and McKenzie2001).
Studies that have focused on either negative or positive
information (forexample Lusk et al. 2001) have shown that negative
information decreasesa consumers’ WTP, while positive information
increases consumers’ WTP.For example, the experimental results of
Lusk et al. (2001) showed apremium involving positive information
regarding the tenderness ofsteak. That is, consumers were willing
to pay a premium when given ataste test that included information
concerning the tenderness of a steakcompared to a taste test with
no information. Results of the study indi-cated that consumers
value quality (as indicated by tenderness in thisstudy) in the beef
market. In another experimental study concerning con-sumer WTP,
Stenger (2000) showed that information resulted in a signifi-cant
increase in WTP for vegetables grown without the use of sewage
1All brand names are described as they were reported in Consumer
Reports. The authors make noclaims to the accuracy of this
information.
Do Consumer Responses to Media Food Safety Information Last?
365
-
sludge as a fertilizer over their current spending levels for
fruits and vege-tables, despite consumers’ perception that the
health risk from such a fer-tilizer treatment was small. Saghaian’s
(2007) observational studyconcerning BSE food safety shocks also
shows that negative informationaffects prices downward over a
ten-week period. In this study, retailprices lag from a week to ten
days behind producer price drops and areless severe.
When consumers receive information about food safety, this
informationis often mixed with both positive and negative messages.
Several studieshave found that negative information has a stronger
effect than positiveinformation. For example, Fox, Hayes, and
Shogren’s (2002) study of irra-diated pork tested favorable versus
unfavorable information to determinewhich had a stronger effect on
consumers’ WTP at the time of purchase.Using three treatments
(negative information, positive information, and abalanced
treatment involving both positive and negative information),
thestudy demonstrated that providing negative information had a
muchstronger influence, even when the negative information comes
from a non-scientific source. In a study using real-life case
reports concerning geneti-cally modified foods, Hu, Zhong, and Ding
(2006) found that positiveinformation did not significantly affect
participants’ WTP, but negativeinformation significantly decreased
WTP.
Several studies have also considered consumers’ response to food
safetyinformation over time. Brown, Cranfield, and Henson (2005)
found thatWTP drops markedly as participants become more tolerant
of risk; theirresults from a one-session experiment indicated that
the WTP of consum-ers who initially overvalued the risk of
food-borne illness tended toincrease their WTP as their tolerance
for risk increased. The study did notcontain a longer-term time
component to determine if participantsreturned to previous levels
of tolerance for risk over time. Hammitt andHaninger’s (2007)
stated-interest survey results indicate that participants’WTP is
insensitive to duration between one and seven days. Shogren,
Listand Hayes (2000) conducted an experimental study involving
consecutiveauctions eliciting participants’ WTP for food products
over a two weekperiod. The authors found that participants’ WTP
changed over time.
Given the strong effect that information has regarding immediate
con-sumers’ food purchasing behavior, food safety concerns can be
expectedto change behavior over time. Studies that have used
observational meth-odologies have indicated that consumer behavior
is affected by thepassage of time. Beach et al. (2008) demonstrated
in their study of theinfluence of newspaper stories about avian
influenza on Italian poultrysales that negative media information
had a persistent effect lasting up tofive weeks. The reasons for
these variations in the duration of impacts onconsumer behavior are
difficult to unravel. Saghaian’s (2007) observationalstudy of the
effect of BSE on U.S. markets found downward price effectsduring
the ten-week study period, forecasted up to 15 weeks.
Thomsen,Shiptsova, and Hamm (2006) found that sales of a recalled
brandremained depressed for approximately eight to twelve weeks and
did notfully recover for four to five months. These results were
specific to aparticular brand that was affected by a recall and did
not pertain tonon-branded food commodity markets.
Applied Economic Perspectives and Policy
366
-
Experimental Design
In empirical studies, a key concern is whether all consumers
have thesame access to information regarding the safety of food
products.Additionally, little is known about how consumer behavior
changes asinformation concerning the incident becomes less
available over time. Tohelp control for these influences, this
study used an experimental econom-ics lab setting. To date,
essentially all experimental studies have looked atconsumer
behavior response to food safety information as captured in asingle
session, and have not conducted the more difficult task of
repeat-edly bringing adult subjects to a lab setting to evaluate
change in WTPover time.
Our experimental design used within-subject comparisons of WTP
fortwo different types of chicken breasts and between-subject
comparisons ofWTP for two different treatment groups – one with
food safety informa-tion and one without this information. This
design was structured to testthe following five primary research
questions in order to understand howpositive and negative
information affected participants in both the short-and
longer-term:
(1) Did consumers have different preferences for the two chicken
typeswhen food safety information was not provided?
(2) Did consumers have different preferences for the two chicken
typeswhen food safety information was provided?
(3) Did brand-related food safety information decrease
consumers’ WTPfor a relatively less safe product?
(4) Did brand-related food safety information increase
consumers’ WTPfor a relatively safer product?
(5) Did consumers’ WTP for any type of chicken in either
treatmentgroup change over time after no additional food safety
informationwas provided?
In this study, 110 adults participated in the experiment.2
Participants wererecruited from a large northeast university’s
lifelong learning campus (forstudents aged 50 and over), as well as
from the university’s staff, andpublic attendees of an annual event
open to the general public. The latterexperiments were conducted in
the Experimental Economics Laboratoryfor Policy and Decision
Research at the University of Delaware. The studywas widely
advertised as an experiment in decision-making that wouldrequire
participation in three sessions. These respective settings
werechosen because they provided an easily accessible laboratory
setting forparticipants that were available in the same building,
at the same timeand day of the week for several weeks. All
experimental sessions wereconducted between April and July 2008.
The subject pool was not selectedto represent the entire United
States or even a regional area.3As shownin table 1, to examine
short- and longer-term impacts on WTP, the
2Initially, 119 subjects participated in the first session, nine
of whom dropped out of the study.Therefore, 92.4% of the subjects
who attended the first session attended all three sessions.3The
sample population used in this study had many participants over the
age of 50. The administra-tors made every effort to ensure that all
participants were fully capable of participating through
severalrounds of instructions. It was not within the scope of this
study to determine if older participants wereless capable of
operating computers, read more carefully, were more risk averse,
were more likely to beexposed to food safety concerns, or were more
rigid in their beliefs. These behaviors may be possiblewith an
older population.
Do Consumer Responses to Media Food Safety Information Last?
367
-
experiment consisted of three sessions. In the first session,
which lastedone hour, participants’ WTP was elicited for each of
four products. Thiswas followed by a second 15-minute session held
at the same time andplace one week after the first session. No
further food safety informationwas given to the participants at the
second session. Participants repeatedpart B of the experiment,
which involved bidding on the four products.The third session was
held 28 days later for half of the participants and 49days later
for the other half.4 This division was necessary to accommodate
Table 1 Experiment structure
Session 1 (1 hour)Part A
1. Participants read instructions
2. Verbal instructions
3. Practice round; WTP bid for pen
Part B
1. Participants read instructions
2. Verbal instructions
3. Presentation of food safety information if session is
Treatment group
4. Presentation of products to participants
5. WTP bids for leading-brand chicken, Ranger brand chicken,
fettuccine pasta,and one dozen eggs
Session 2 (7 days later, 15 minutes)Part B only, no additional
information
1. Participants read instructions
2. Verbal instructions
3. WTP bids for leading-brand chicken, Ranger brand chicken,
fettuccine pasta,and one dozen eggs
Session 3 (28 days or 49 days later, 15 minutes)Part B only, no
additional information
1. Participants read instructions
2. Verbal instructions
3. WTP bids for leading-brand chicken, Ranger brand chicken,
fettuccine pasta,and one dozen eggs
4A limiting factor for the length of the study was the semester
length at the lifelong learning facility,which ended eight weeks
after the beginning of this study. Thus, extending the study beyond
sevenweeks would have likely meant a much lower percentage of
participants completing all three sessions.
Applied Economic Perspectives and Policy
368
-
scheduling conflicts, such as a national holiday. Both
treatments wereequally represented in the different lengths of time
for the third sessions.In both treatments, the third session lasted
fifteen minutes, and Part B wasrepeated with no further food safety
information presented.
While 119 subjects initially participated in the first session,
nine of thesesubjects subsequently dropped out of the study.
Therefore, 92.4% of thesubjects who attended the first session
attended all three sessions.Participants earned $60 in cash and/or
products for the experiment. Thepayments for all participants were
$11 in cash and/or products insession 1, $10 in cash and/or
products in session 2, and the remainingpayment in cash and/or
products in session 3.
A modification of the sealed-bid English auction mechanism was
used toelicit WTP estimates due to its ease of explanation and its
demand-revealing and incentive-compatible properties (Davis and
Holt 1993;Bernard 2006). Using this auction mechanism, subjects’
incentives were tobid their highest WTP and therefore reveal their
true demand preferences.Compared to second-price auctions,
traditional English auctions are betterable to measure
participants’ WTP, but have the disadvantage of havingbids visible
to everyone in the experiment session, which can lead to poten-tial
group-effects on individual behavior. Lusk and Shogren’s (2007)
reviewof different auction methods also point out the merits of the
Englishauction, as well as the random nth price auction. The
sealed-bid Englishauction retains the benefit of the increasing bid
clock while keeping bidshidden (Bernard, 2006).
In this study, we used a modification of the traditional English
Auction.We refer to this auction as a sealed, random nth bid
English Auction. All ofthe bids were submitted confidentially so
that participants were unaware ofwhat other participants in their
experimental session were bidding. Thisresearch used $10 as the
maximum bid since this was the amount of theinitial balance given
to all participants and was significantly higher thanthe market
price for 1.5 pounds of chicken breasts. In these experiments,the
highest bid submitted was $9.90 – for Ranger chicken.
Participants were informed using both written and oral
instructions thatthe optimal strategy was to bid their actual
highest WTP for each product.Therefore, they needed to determine
their personal highest WTP for theproduct being auctioned and
submit that amount as their bid. A computerprogram that employed
Excel spreadsheets and programmed using VisualBasic for
Applications was used to confidentially record participants’WTP
(Messer, Kaiser, and Schulze, 2008). Participants were instructed
tostop the program when the displayed price reached her/his
maximumWTP. The program initially showed a $0.00 price. Subjects
wanting tosubmit a bid of $0 were provided a “Withdraw” button at
this initialscreen. Participants wanting to submit bids greater
than zero wereinstructed to start by clicking the “Start” button.
When they did so, theprogram began increasing the price in one-cent
increments at a uniformtime interval until the participant clicked
a button marked “Withdraw.” Ifthe “Withdraw” button was never
clicked, the program would record themaximum bid allowed ($1 for
the training round and $10 for the fourproducts). The maximum price
was always equal to the participants’initial balance of funds. To
help ensure that participants did not get asense of what other
participants were bidding, it was necessary to allowthe clock to
run up to the maximum bid of $10.
Do Consumer Responses to Media Food Safety Information Last?
369
-
The price for the item was determined by the highest rejected
bid. At theend of each auction, bids were arranged by the
administrator’s computerprogram from highest to lowest and the
number of purchasers was ran-domly determined. The subject with the
highest bid was the first pur-chaser, and so on. The number of
purchasers was determined randomlyby having a volunteer roll a
six-sided die with the number of purchaserstherefore ranging from
one to six for sessions with seven or moreparticipants.5 For
example, if the number of participants in a session wasten, and the
result of the die roll was five, the five highest bidders
wouldpurchase the product and pay a price equal to the bid of the
sixth highestbidder.
Prior to conducting the experiments, participants were
randomlyassigned to either a treatment group (n ¼ 56) that received
media-basedfood safety information, or a control group (n ¼ 54)
that received nomedia information. The average number of
participants in a session wasnine, with a range from three to
sixteen. Verbal and oral instructions wereprovided to improve the
understanding of experiment procedures (see theReview Appendix).
Questions were encouraged during all stages of theexperiment.
Subjects were seated at individual computer terminalsequipped with
privacy screens so that all decisions would be
madeconfidentially.
The first session was divided into two parts (table 1).
Following Messeret al. (2011), Part A consisted of a practice round
where participants bidon a pen. This practice round helped ensure
that participants understoodhow the bidding process worked and how
to stop the computer programat the desired WTP. For the pen, the
initial balance provided to eachsubject was $1. Participants were
instructed to bid zero if they valued theproduct at $0 or less, and
were instructed to bid $1 if they valued theproduct at $1 or
greater. If their value for the pen was between $0 and $1,they were
instructed to stop the computer program at the price that
repre-sented their highest WTP. During the training round,
questions wereanswered to ensure that participants understood the
program andprocedure.
In part B, the initial balance was $10 for each product.
Participants werepermitted to bid between $0 and $10. In part B,
four products wereauctioned:
1) Frozen boneless skinless chicken breasts from a leading-brand
– suchas Foster Farms, Perdue, Pilgrim’s Pride, or Tyson
(approximately112 pounds).
2) Frozen boneless skinless chicken breasts from Ranger
(approximately112 pounds).
6
5For sessions with less than seven participants, the maximum
number of purchasers was k – 1, wherek was the number of
participants. In some cases, a four-sided die was used. The number
of purchaseswas determined randomly using an nth-price auction
because of the repeated nature of these experi-ments. For instance,
if a standard second-price auction was used, then after the first
session, most sub-jects would have a good sense of whether their
bid for the selected product in the subsequent sessionswould be
close to the highest bid in their group. The use of an nth-price
auction helped ensure thatessentially none of the bidders would be
“off-margin.”6Ranger brand chicken is only available in the Pacific
Northwest. To avoid deceiving subjects, over 70pounds (nearly 50
packages) of Ranger brand frozen boneless skinless chicken breast
were shipped over-night express from Bellingham, Washington,
packaged in dry ice and in freezer containers, at a cost ofover a
thousand dollars.
Applied Economic Perspectives and Policy
370
-
3) Eggs (one dozen, size large).4) Fettuccine pasta (one
pound).
All original packaging information was removed from the products
priorto the experiment to control for any reaction individuals
might have to thepackaging. Both types of chicken were displayed in
clear, one-gallonsealed freezer bags. Participants were informed
that if they purchased thechicken, we would package the chicken in
ice, if desired, to ensure thatthe chicken would remain frozen
until they returned home.7 Before thebidding began, an
administrator walked around the room and displayedeach of the
products. Participants were asked not to touch the products,since
the products would be distributed after the experiment to the
pur-chasers, but they were permitted to visually inspect each of
them asclosely as they wished. To prevent potential order effects,
the order inwhich the products were displayed was determined using
a Latin squaresdesign.8
Participants bid on all four products, but only one product was
actuallypurchased similar to the procedures designed by Bernard,
Zhang, andGifford (2006). The purchased product was randomly
predeterminedusing a four-sided die and was written on an index
card. The card wassealed inside an envelope, which was opened by a
volunteer after all ofthe auctions were completed. Prior studies
have shown that, in amultiple-round auction such as this one,
randomly determining theproduct that is purchased helps to elicit
WTP among participants (seeLusk, Feldkamp, and Schroeder (2004);
Hayes et al. (1995); and Messeret al. (2010)). This procedure
compensates for the potential that a partici-pant purchased a
product in one session and therefore decreased her orhis WTP in
subsequent sessions. It was possible for the same product tobe
binding in more than one session. The binding product in this
experi-ment was randomly determined for each session in an effort
to reduce thepotential effects across sessions caused by a subject
receiving a product inone session and having to bid again for the
same product in subsequentsessions.
These procedures were used since this experiment involved
multiplesessions with the same products. In such cases, there is a
potential thatparticipants who purchase a product in one session
will bid lower thantheir true WTP in subsequent sessions since they
already have theproduct. However, this effect was minimized since
the actual quantity ofchicken sold during the experiments was
small. Therefore, it was impor-tant to make the binding auctions
random in order to minimize this typeof effect on participant
WTP.
After bids for all four products were collected, the
predeterminedproduct was distributed to purchasers and used to
calculate cash earnings.
7All of the chicken was frozen so that consumers would have less
concern about potential food safetyissues related to the
administrator’s handling of the chicken, and to minimize concerns
that the chickenwould need to be eaten immediately. This latter
reason was important in the case that participantswere not
returning to their homes immediately after the research, as well as
to ensure that the potentialuseful life of this product was as long
as possible, so that consumers WTP would not be
significantlyaffected by the amount of chicken that they might have
just recently purchased at a grocery store priorto the session.8A
Latin Squares design is used here as a method of varying the order
of products. For instance, if inthe first session, products A, B,
and C are introduced in that order, in the next session, they will
beintroduced in the order B, C, A. In the following session they
will be in the order C, A, B, and so on.
Do Consumer Responses to Media Food Safety Information Last?
371
-
At the end of the experiment a short survey was then distributed
in eachsession to collect demographic data, as well as information
about subjects’beliefs about the safety of food products, and
whether subjects receivedadditional food safety information between
sessions of the experiment.The information provided to the
treatment group about the two chickenproducts is included in
appendix 1.
Results
The participants’ WTP bids for chicken are displayed in the four
panelsof figure 1, which shows the demand for each chicken type in
each sessionand in each treatment group. The figures plot WTP on
the y-axis againstthe percentage of participants willing to pay at
a given price on the x-axis.
Figure 1 WTP for approximately 1 12 pounds of chicken breasts,
by type and treatment
Applied Economic Perspectives and Policy
372
-
The legends of each figure provide the mean and standard
deviationrelated to each demand curve.
To answer the research questions outlined in the experimental
designsection, the participants’ responses were analyzed in several
differentways. Normality tests were performed on WTP data for the
leading-brandchicken and the Ranger brand of chicken using STATA’s
sktest, whichmeasures skewness and kurtosis to test for normality.
The results of theskewness and kurtosis tests indicated that the
data were not normally dis-tributed. Thus, the WTP bids were
analyzed with nonparametricWilcoxon rank-sum tests to analyze
differences between the two treatmentgroups (table 2). Wilcoxon
signed-rank tests were used to test for statisticalsignificance
within experiment sessions and within chicken types.
Additionally, two-limit Tobit regression models were also used
toexplain participants’ WTP for chicken. A two-limit Tobit was
required,since participants’ bids were constrained at $0 and $10 in
the experiment.The model included data from all three sessions, for
both chicken typesfrom both the control and treatment groups. This
model thus included sixobservations from each of the 103
non-vegetarian subjects bidding threetimes each for two brands of
chicken. A random-effects model was usedsince each participant had
multiple observations in the panel data set.Marginal-effect
coefficients were calculated to translate the
regressioncoefficients into WTP. The model included an information
variable meas-uring the impact of the food safety information in
the study on partici-pants’ WTP. The variable info was also
interacted with all otherindependent variables to find the effects
of food safety information onsubgroups. Table 3 provides
descriptive statistics on the variables in thefollowing model:
wtp_chicken ¼ b0 + b1(info) + b2(ranger) + b3(ranger*info) +
b4(days) + b5(days2) +b6(days*info) + b7(days2*info) +
b8(ranger*days) + b9(ranger*days2) +b10(ranger*info*days) +
b11(ranger*info*days2) + b12(age) + b13(age*info) +b14(female) +
b15(female*info) + b16(children) + b17(children*info)
+b18(primary_shopper) + b19(primary_shopper*info) + b20(education)
+b21(education*info) + b22(nonwhite)+ b23(nonwhite*info)+
b24(p_chickensafe) +b25(p_chickensafe*info) + b26(income)+
b27(income*info)+
b28(eat_chicken_often)+b29(eat_chicken_often*info) + 1.
The variables p_chickensafe and p_chickensafe*info were
constructed usingsurvey responses from participants indicating
whether they consideredchicken to be a safe product. A Hausman test
for endogeneity indicatedthat the p_chickensafe variable was
endogenous. The residual term was a sig-nificant predictor of the
dependent variable (p ¼ 0.023). Therefore, instru-mental variables
were constructed for p_chickensafe and p_chickensafe*infousing
auxiliary regressions of these variables on all exogenous variables
inmodel.
Tests of Research Questions
The first research question was, “Did consumers have different
preferencesfor the two chicken types when food safety information
was not provided?” Avisual inspection of panels A and C in figure 1
suggests that the demandcurves for leading-brand chicken were
slightly higher than the demandcurves for the Ranger brand. As can
be seen in table 2, panel A, onaverage in the three sessions,
participants were willing to pay between
Do Consumer Responses to Media Food Safety Information Last?
373
-
Table 2 Sets of tests of differences in WTP with Wilcoxon
significance tests.
H0: WTP for Ranger brand - WTP for Leading-brand 5 0
Panel Control Prob> |z| Panel Treatment Prob> |z|
A Session 1 2.57 – 2.86 ¼ 20.29 0.1018 B 3.15 – 2.21 ¼ 0.94
0.000Session 2 2.37 – 2.61 ¼ 20.24 0.2059 2.41 – 2.05 ¼ 0.36
0.036Session 3 2.38 – 2.53 ¼ 20.15 0.3040 2.32 – 1.95 ¼ 0.37
0.012All Sessions 2.44 – 2.67 ¼ 20.23 NA 2.62 – 2.07 ¼ 0.55 NA
H0: WTP for Treatment group – WTP for Control group ¼
0Leading-brand Prob. |z| Ranger brand Prob. |z|
C Session 1 2.21 – 2.86 ¼ 20.65 0.0219 D 3.15 – 2.57 ¼ 0.58
0.2821Session 2 2.05 – 2.61 ¼ 20.56 0.0929 2.41 – 2.37 ¼ 0.03
0.8146Session 3 1.95 – 2.53 ¼ 20.58 0.0691 2.32 – 2.38 ¼ 20.06
0.8247All Sessions 2.07 – 2.67 ¼ 20.60 NA 2.62 – 2.44 ¼ 0.18 NA
H0: WTP for Session 1 – WTP for Session 2 ¼ 0Leading-brand Prob.
|z| Ranger brand Prob. |z|
E Control 2.86 – 2.61 ¼ 0.25 0.1253 F 2.57 – 2.37 ¼ 0.20
0.2159Treatment 2.21 – 2.05 ¼ 0.16 0.6146 3.15 – 2.41 ¼ 0.76
0.0018
H0: WTP for Session 2 – WTP for Session 3 ¼ 0Leading-brand Prob.
|z| Ranger brand Prob. |z|
G Control 2.61 – 2.53 ¼ 0.08 0.7252 H 2.37 – 2.38 ¼ 20.01
0.8256Treatment 2.05 – 1.95 ¼ 0.10 0.8156 2.41 – 2.32 ¼ 0.09
0.9818
H0: WTP for Session 1 – WTP for Session 3 ¼ 0Leading-brand Prob.
|z| Ranger brand Prob. |z|
I Control 2.86 – 2.53 ¼ 0.33 0.1424 J 2.57 – 2.38 ¼ 0.21
0.2432Treatment 2.21 – 1.95 ¼ 0.26 0.8982 3.15 – 2.32 ¼ 0.83
0.0171
Applied
Econ
omic
Perspectives
and
Policy
374
-
$0.15 and $0.29 less for a pound and a half of the chicken
breasts.However, these differences were not statistically
significant in any session.For the pooled data used in the
two-limit Tobit analysis, the coefficient onthe variable ranger was
again negative (-0.273) and not statistically signifi-cant (p ¼
0.216) (table 4).
Recall that participants were given information which stated
thatRanger brand was safer. The second question, “Did consumers
have different
Table 3 Variables in the Tobit model.
Variable MeanStandarddeviation Minimum Maximum
wtp_chicken - WTP for bothleading-brand and Rangerchicken.
2.430 1.787 0.0 9.9
info - 1 for the treatment group and0 otherwise.
0.509 0.500 0.0 1.0
ranger - 1 for WTP for Ranger and 0otherwise.
0.500 0.500 0.0 1.0
days - Number of days since thefirst session.
15.024 17.745 0.0 49.0
days2- Number of days since thefirst session squared.
540.083 860.511 0.0 2401.0
age- 19 if , 20; 24.5 if between 20and 29; 34.5 if between 30
and39, . . .; 84.5 if between 80 and 89.
47.211 16.233 19.0 84.5
female- 1 for female and 0otherwise.
0.668 0.471 0.0 1.0
children- 1 if participant haschildren under 18 living at
homeand 0 otherwise.
0.187 0.390 0.0 1.0
primary_shopper - 1 if participant isprimary shopper in
householdand 0 otherwise.
0.730 0.444 0.0 1.0
education- 0 if less than collegedegree, 1 if college degree,
and 2if more than a college degree.
1.332 0.812 0.0 2.0
nonwhite - 1 if nonwhite racialdesignation and 0 otherwise.
0.104 0.305 0.0 1.0
p_chickensafe - Instrumental variableconstructed from
participantresponses whether they considerchicken safe.
2.746 0.574 1.0 4.2
income- Categorical householdannual income variable rangingfrom
0 ($0 - $39,999) to 10 (over$200,000) by $40,000 increments.
73.772 54.500 10.0 240.0
eat_chicken_often - 0 if participanteats chicken never; 1 if
rarely; 2 ifonce a month; 3 if several times amonth; 4 if once a
week; and 5 ifseveral times a week.
2.713 1.205 0.0 4.0
Do Consumer Responses to Media Food Safety Information Last?
375
-
preferences for the two chicken types when food safety
information was provided,”contrasted WTP in the information
treatment for the two chicken types.Inspection of figure 1 shows a
marked increase in WTP betweenleading-brand chicken (panel D) and
Ranger chicken (panel B). These dif-ferences in average WTP ranged
from $0.36 to $0.94, which were statisti-cally significant in
session 1 (p ¼ 0.000), session 2 (p ¼ 0.036), and session3 (p ¼
0.012) (table 2, panel B). This result was further supported by
theTobit model results where the increase in WTP for Ranger from
the foodsafety information was $1.24 when all other factors were
held constant
Table 4 Two-limit Tobit results, WTP for chicken.
Marginal effect Standard error P> |z|
constant 21.526 1.676 0.362info 1.236 2.079 0.552ranger 20.273
0.221 0.216ranger*info 1.234 0.318 0.000days 20.037 0.023
0.116days2 0.001 0.001 0.161days*info 20.050 0.036 0.170days2*info
0.001 0.001 0.137ranger*days 20.008 0.031 0.797ranger*days2 0.000
0.001 0.696ranger*info*days 0.010 0.048 0.843ranger*info*days2
20.001 0.001 0.578age 0.015 0.016 0.370age*info 20.012 0.022
0.595female 20.061 0.505 0.904female*info 0.169 0.674 0.802children
0.926 0.513 0.071children*info 22.551 0.854 0.003primary_shopper
1.053 0.587 0.073primary_shopper*info 20.103 0.777 0.894education
0.216 0.293 0.461education*info 20.648 0.392 0.098nonwhite 20.066
0.575 0.908nonwhite*info 21.113 1.014 0.272p_chickensafe 0.956
0.383 0.013p_chickensafe*info 20.967 0.496 0.051income 0.007 0.004
0.095income*info 20.006 0.006 0.360eat_chicken_often 20.309 0.196
0.116eat_chicken_often*info 1.157 0.266 0.000Sum of all info
coefficients 21.639F test statistic 8.550Wald chi2 ¼ 118.820Prob.
chi2 ¼ 0.000Log likelihood ¼ 2863.063Left-censored observations
79Uncensored observations 455Right-censored observations 0
Note: Since some of the survey questions were left blank by
subjects, some observations were notincluded in the final model,
reducing the total number of observations to 534.
Applied Economic Perspectives and Policy
376
-
(p ¼ 0.000). All of this evidence suggested that the null
hypothesis can berejected, as participants in the treatment group
were willing to pay morefor Ranger chicken than for leading-brand
chicken. This response to thepositive information for Ranger
chicken9 corroborated results found inLusk et al. (2001) and
Stenger (2000).
The third research question, “Did brand-related food safety
informationdecrease consumers’ WTP for a relatively less safe
product,” related to whetherinformation affects WTP for chicken.
Since all variables were interactedwith the information variable in
the Tobit analysis, the derivative of infowas calculated by adding
the coefficient for info and all variables inter-acted with info. A
model-specification F-test indicated that the info variablehad a
coefficient of -1.639 and was significant at the p , 0.01
significancelevel (table 4). The effect of information was most
pronounced in thelowering of WTP for leading-brand chicken. For
instance, as shown inpanel C of table 2, participants who were
given food safety informationfrom Consumer Reports were willing to
pay, on average, $0.65 less in thefirst session for leading-brand
chicken than participants who were notgiven this information. This
difference was statistically significant(p ¼ 0.0219) (table 2,
panel C). The $0.56 and $0.58 average differences inthe second (p ¼
0.0929) and third (p ¼ 0.0691) sessions, respectively, werenot
significant at the p ≤ .05 significance level, but were significant
at the0.10 level. Taken together, this suggested that the answer to
the thirdresearch question was that consumers do reduce their WTP
for the lesssafe product after receiving negative food safety news.
The results forleading-brand chicken were consistent with the
literature (see Hayes et al.1995 and Messer et al. 2011) in that
negative information decreased WTP.
Research question 4, “Did brand-related food safety information
increase con-sumers’ WTP for a relatively safer product,” reveals
the result that food safetyinformation for Ranger chicken had a
significant positive effect. Table 4indicates that the Tobit model
results for the variable ranger*info arehighly significant with a
coefficient of 1.234. All else held constant,average WTP for Ranger
chicken was $1.23 higher. Table 2 (panel B) alsoindicated that
participants who received food safety information werewilling to
pay a premium for Ranger chicken when compared to leading-brand
chicken in all three sessions. In contrast, the information
treatmentdid not increase participants’ WTP for Ranger chicken by
session, as noneof the differences between any of the three
sessions were statistically sig-nificant (table 2, panel D). Thus,
consumers will pay more for a saferproduct after learning of
positive food safety information. Furthermore,consumers’ WTP for
this product will not decrease like it does for the lesssafe brand
of chicken. However, this result is not differentiable by
session,which suggests that time is not a factor in consumer
willingness to pay apremium for a safer product.
The fifth research question, “Did consumers’ WTP for any type of
chickenin either treatment group change over time after no
additional food safety infor-mation was provided,” tested the
longer-term effect on WTP. As shown infigure 1, average WTP for
both types of chicken in both treatment groupswas highest in the
first session. However, tests of these differences
9While the information about Ranger brand chicken was not
universally positive, as some level of bac-teria contamination was
detected, the Consumer Reports article clearly presented Ranger as
the rela-tively safer alternative for chicken consumers.
Do Consumer Responses to Media Food Safety Information Last?
377
-
(table 2, panels E-J), show that the only differences that were
statisticallysignificant were the $0.76 and $0.83 decrease in WTP
for Ranger chickenin the information treatment (p ¼ 0.0018 and p ¼
0.0171, respectively).The two-limit Tobit model indicates that none
of the eight variables thatincluded days were statistically
significant. The variable, days, measuredthe number of days that
had elapsed since the first session. This passageof time could
potentially have affected participants in several ways:
partic-ipants may have forgotten about the information they learned
over time;participants’ desire for chicken may also have changed
over time based onwhether they had recently purchased chicken;
participants may also havechanged their mind or had second thoughts
since the previous sessions.The insignificance of the days
variables demonstrates that there was no sig-nificant change in
participants’ WTP over time for leading-brand orRanger brand
chicken in either treatment group.10
A more differentiated analysis is available from the Wilcoxon
signifi-cance tests shown in table 2. The results of these tests
indicate that WTPfor Ranger was significantly different between
sessions 1 and 2 in thetreatment group (panel F). Session 1 WTP for
Ranger is also significantlydifferent from session 3 for the
treatment group (panel J). However, WTPfor Ranger is not
significantly different between sessions 2 and 3 in thetreatment
group (panel H). None of the differences in WTP for Rangerbetween
any of the sessions is significantly different in the control
group.WTP for leading-brand is not significantly different from
session tosession for either the treatment or the control group
(panels E, G, and I).These results indicate that the premium for
Ranger is short-lived and dis-appears by session 2, while the
negative effect on WTP for leading-brandremains over all
sessions.
In addition to treatment dummy variables, several demographic
varia-bles were also included in the Tobit regression model. As is
common inexperimental studies, gender, age, income, and education
variables wereincluded as explanatory variables of WTP (Bernard,
Zhang, and Gifford2006; Fox, Hayes, and Shogren 2002; Hobbs et al.
2005; Hu, Zhong, andDing 2006; Lusk et al. 2001; Lusk, Feldkamp,
and Schroeder 2004). Sincesome research has found the presence of
children in participants’ house-holds to influence WTP, the number
of children in each household wasincluded as an explanatory
variable (Bernard, Zhang, and Gifford 2006;Hu, Zhong, and Ding
2006; Kanter, Messer, and Kaiser 2009; Lusk et al.2001; Messer et
al. 2011). Following previous studies, race (Bernard,Zhang, and
Gifford 2006), whether a subject considers themselves theprimary
shopper in the household (Kanter, Messer, and Kaiser 2009), and
10Survey results indicate that 24 subjects (21.8% of the total
sample) reported receiving outside infor-mation between
experimental sessions that affected their bidding during the course
of the experiment.Of these 24 subjects, 9 were in the control group
and 15 were in the treatment group. While subjectswere not given
any information about Ranger brand chicken other than the safety
information providedin the Consumer Reports article, simple
internet searches about this brand would reveal to consum-ers that
Ranger is also free range and the chickens are raised without the
use of hormones or antibiot-ics. To test whether subjects
researched Ranger brand chicken and subsequently bid higher
insubsequent sessions, any positive responses were compared with
bids that were at least 10% higher forRanger brand chicken. Of the
subjects who responded that they were affected by outside
information,only 4.6% of subjects bid at least 10% higher for
Ranger in session two than in session one, 6.4% ofsubjects bid at
least 10% higher in session three than in session two, and 5.5% bid
at least 10%higher in session three than in session one. Therefore,
the potential impact of information gainedoutside of the experiment
is likely minimal.
Applied Economic Perspectives and Policy
378
-
the frequency which subjects reported that they eat chicken
(Messer et al.2011) were also included as explanatory
variables.
Many results from the above mentioned variables are worth
noting. In theTobit model, children*info, p_chickensafe, and
eat_chicken_often*info were highlysignificant independent
variables. The variables children, primary_shopper,
edu-cation*info, p_chickensafe*info, and income were also
significant. Participants inthe treatment group with children
(children*info) under 18 living at home werewilling to pay $2.55
less on average for chicken. Those in the control group(children)
were willing to pay $0.93 more, on average. Those participants in
thetreatment group who considered chicken to be a safe product were
willing topay $0.96 more on average. Interestingly, those
participants in the treatmentgroup who ate chicken often
(eat_chicken_often*info) were willing to pay $1.16more on average.
This result was consistent with the results of Payne et al.(2009),
who found frequent consumers of beef were less affected by
negativefood safety information. Participants who consume chicken
more frequentlyare also more likely to have been exposed to similar
food safety warnings con-cerning chicken in the past. This may also
be a factor in the lack of significanceof this variable for the
control group and such a strongly significant result forthe
treatment group. The significance of the variable primary_shopper
indicatedthat participants in the control group who considered
themselves to be theprimary shopper in the household were willing
to pay $1.05 more on averagethan those who did not consider
themselves to be the primary shopper. Thesignificance of the impact
of product safety information and education (educa-tion*info)
indicated that more highly educated participants in the
treatmentgroup were willing to pay, on average, $0.43 less ($0.65
less + $0.22 from thecoefficient for the education variable). The
significance of the income variable(income) indicated that for
every dollar of household income that participantsin the control
group reported, participants were willing to pay $0.007 more
forchicken. The remaining demographic variables were statistically
insignificant.
Conclusion
This experimental study demonstrated that a combination of
negativeand positive information regarding food safety has a
long-lasting impacton demand. One hundred ten adults participated
in this research to testthe effects of negative and positive
information over time on consumerpurchasing behavior (as measured
by WTP). Participants were asked torepeat the experiment twice
after the initial session to measure changes intheir WTP over time,
extending out to seven weeks. During each session,WTP data were
collected on both leading-brand and Ranger brandchicken.
Demographic data were also obtained from the participants. Thefood
safety information used in the study came from a 2007
ConsumerReports magazine article stating that leading-brand
chicken, specificallyidentified as Perdue, Tyson, Foster Farms, and
Pilgrim’s Pride, frequentlycontained harmful bacteria. The article
also stated that another brand,Ranger, was relatively free of
harmful bacteria.
Results from this study indicated that consumers are willing to
changetheir purchasing behaviors to avoid unsafe products. Both
positive andnegative information had an effect on consumers’ WTP.
Consumers werewilling to pay less for the leading-brand chicken
after they received nega-tive food safety information compared to a
control group that did not
Do Consumer Responses to Media Food Safety Information Last?
379
-
receive this information. Participants that received positive
food safetyinformation about Ranger brand chicken were consistently
willing to paymore for this safer option than for leading-brand
chicken. This suggeststhat when the information about which brands
are safer is available, con-sumers are willing to alter their
purchasing behavior to favor the saferalternative, even if it was a
relatively unknown brand. Both of theseeffects appeared to last
well beyond the initial exposure to mediainformation.
The results of this study provided limited guidance for food
industriesaiming to avoid potential losses from media reports
regarding food safetyconcerns. Like consumers, an industry may be
affected severely by mediareports about food safety that lead to a
long-lasting decrease in consumerwillingness to buy the affected
product. Negative information about a spe-cific brand or specific
brands can, over time, affect an entire industry. Thiscan be
especially devastating to a seasonal agricultural industry, an
indus-try that has many food substitutes, or in situations where
the true sourceof the food contamination is incorrectly identified.
Results from this studyindicated that introducing safer alternative
products when feasible canhelp capture the premiums that are lost
during an incident that raisesfood safety concerns. Unlike some of
the event studies (see Thomsen andMcKenzie 2001; Lusk and Schroeder
2002), which focus on the observatio-nal effects of negative food
safety information only, this study was able toobserve a measurable
difference between a relatively more safe and rela-tively less safe
product over time. The measurable difference betweenproducts is
short-lived, while the negative effect of food safety informationon
the less safe product persists in the longer term. The
longer-termdecrease in consumer willingness to pay for
leading-brand chicken sug-gests that negatively impacted food
products will result in significantlylower consumer WTP for an
extended time. This suggests that industryshould seek to prevent
food safety problems, or face an extended decreasein consumer
demand.
An obvious limitation of this study is its seven-week duration.
Thislength was initially chosen to align with several empirical
studies on con-sumer responses to food safety concerns, and also
due to logistical con-straints on recruiting adult subjects for a
repeated experiment. Given theresults suggesting that consumer WTP
was still affected upwards of sevenweeks, investigations into a
longer duration are warranted. However, thisseven-week study is to
our knowledge the longest of its type to date.There is a need for
longer-term experimental studies to overcome the pos-sibility that
experimental studies merely capture a snapshot of subjects’behavior
and are not an accurate portrayal of their true behavior overtime.
Another potential limitation of the experimental setting over time
inthis study is the possibility that subjects interacted and
affected eachothers’ bidding behavior between sessions, or that
subjects were influ-enced by other sources of information between
sessions. This study triedto address these concerns by using a
between-subject control and treat-ment design, and asked all
participants to sign a confidentiality agree-ment11. Further
investigation of these issues will help develop a better
11The confidentiality agreement stated, “I agree that by
participating in this experiment I will notdiscuss or make known
the experiment or any part thereof to any individual.”
Applied Economic Perspectives and Policy
380
-
understanding of how long consumers’ responses are influenced by
mediafood safety information.
Supplementary Material
Supplementary material is available at Applied Economic
Perspectives andPolicy online
(http://aepp.oxfordjournals.org/).
References
Beach, Robert H., Fred Kuchler, Ephraim Leibtag, and Chen Zhen.
The Effects ofAvian Influenza News on Consumer Purchasing Behavior:
A Case Study of ItalianConsumers’ Retail Purchases. USDA Economic
Research Service, EconomicResearch Report 65. August 2008.
Bernard, John C. 2006. Finding and Retaining the Dominant
Strategy: TheSecond-price, English, and ‘Sealed Offer’ English
Auctions. Journal of EconomicBehavior and Organization 60:
553–562.
Bernard, J.C., and D. J. Bernard. 2009. What is it about Organic
Milk? AnExperimental Analysis. American Journal of Agricultural
Economics 91: 826–836.
Bernard, John C., Chao Zhang, and Katie Gifford. 2006. An
ExperimentalInvestigation of Consumer Willingness to Pay for Non-GM
Foods When anOrganic Option Is Present. Agricultural and Resource
Economics Review 35(2):374–386.
Brown, Jennifer, John A.L. Cranfield, and Spencer Henson. 2005.
RelatingConsumer Willingness-to-Pay for Food Safety to Risk
Tolerance: AnExperimental Approach. Canadian Journal of
Agricultural Economics 53: 249–264.
Calvin, L. 2007. Outbreak Linked to Spinach Forces Reassessment
of Food SafetyPractices. Amber Waves 5(3): 24–32.
Centers for Disease Control. 2008. Outbreak of Salmonella
Serotype SaintpaulInfections Associated with Multiple Raw Produce
Items—United States.Morbidity and Mortality Weekly Report 57(34):
929–934.
Consumer Reports. 2007. Dirty Birds: Even ‘Premium’ Chickens
Harbor DangerousBacteria. Consumer Reports: 20–24.
Crutchfield, Stephen R., and Tanya Roberts. 2000. Food Safety
Efforts Accelerate inthe 1990s. FoodReview 23(3): 44–50.
Davis, Douglas D., and Charles A. Holt. 1993. Experimental
Economics, Princeton,New Jersey: Princeton University Press.
Fox, John A., Dermot J. Hayes, and Jason F. Shogren. 2002.
Consumer Preferencesfor Food Irradiation: How Favorable and
Unfavorable Descriptions AffectPreferences for Irradiated Pork in
Experimental Auctions. The Journal of Risk andUncertainty 24(1):
75–96.
Hammitt, James K., and Kevin Haninger. 2007. Willingness to Pay
for Food Safety:Sensitivity to Duration and Severity of Illness.
American Journal of AgriculturalEconomics 89(5): 1170–1175.
Hayes, D.J., J.A. Fox, and J.F. Shogren. 2002. Experts and
Activists: HowInformation Affects the Demand for Food Irradiation.
Food Policy 27: 185–193.
Hayes, Dermot J., Jason F. Shogren, SeungYoull Shin, and James
B. Kliebenstein.1995. Valuing Food Safety in Experimental Auction
Markets. American Journal ofAgricultural Economics 77(1):
40–54.
Hu, Wuyang, Funing Zhong, and Yulian Ding. 2006. Actual Media
Reports on GMFoods and Chinese Consumers’ Willingness to Pay for GM
Soybean Oil. Journalof Agricultural and Resource Economics 31(2):
376–391.
Kanter, Christopher, Kent D. Messer, and Harry M. Kaiser. 2009.
Does ProductionLabeling Stigmatize Conventional Milk? American
Journal of AgriculturalEconomics 91(4): 1097–1109.
Do Consumer Responses to Media Food Safety Information Last?
381
http://aepp.oxfordjournals.org/http://aepp.oxfordjournals.org/http://aepp.oxfordjournals.org/http://aepp.oxfordjournals.org/http://aepp.oxfordjournals.org/
-
Lusk, Jayson L., John A. Fox, Ted C. Schroeder, James Mintert,
and MohammadKoohmaraie. 2001. In-Store Valuation of Steak
Tenderness. American Journal ofAgricultural Economics 83:
539–551.
Lusk, Jayson L., Ty Feldkamp, and Ted C. Schroeder. 2004.
Experimental AuctionProcedure: Impact on Valuation of Quality
Differentiated Goods. AmericanJournal of Agricultural Economics
86(2): 389–406.
Lusk, Jayson L., and Ted C. Schroeder. 2002. Effects of Meat
Recalls on FuturesMarket Prices. Agricultural and Resource
Economics Review 31/1 (April): 47–58.
Lusk, Jayson L., and Jason Shogren. Experimental Auctions:
Methods and Applicationsin Economic and Marketing Research.
Cambridge, UK: Cambridge UniversityPress, 2007.
Marsh, Thomas L., Ted C. Schroeder, and James Mintert. 2004.
Impacts of MeatProduct Recalls on Consumer Demand in the USA.
Applied Economics 36:897–909.
Mead, Paul S., Laurence Slutsker, Vance Dietz, Linda F. McCaig,
Joseph S. Bresee,Craig Shapiro, Patricia M. Griffin, and Robert V.
Tauxe. 1999. Food-RelatedIllness and Death in the United States.
Emerging Infectious Diseases 5(5): 607–626.
Messer, Kent D., Harry M. Kaiser, Collin Payne, and Brian
Wansink. 2011. “CanAdvertising Alleviate Consumer Concerns About
Mad Cow Disease?” AppliedEconomics 43(12):1535–1549.
Messer, Kent D., Harry M. Kaiser, and William D. Schulze. 2008.
“The Problem ofFree Riding in Voluntary Generic Advertising:
Parallelism and PossibleSolutions from the Lab” American Journal of
Agricultural Economics 90(2):540–552.
Messer, Kent D., Greg L. Poe, Daniel Rondeau, William D.
Schulze, and ChristianVossler. 2010. “Social Preferences and
Voting: An Exploration Using a NovelPreference Revealing
Mechanism.” Journal of Public Economics 94(3–4): 308–317.
Payne, Collin, Kent D. Messer, and Harry M. Kaiser. 2009. Which
Consumers AreMost Responsive to Media Induced Food Scares?
Agricultural and ResourceEconomics Review 38(3): 295–310.
Piggott, Nicholas E., and Thomas L. Marsh. 2004. Does Food
Safety InformationImpact U.S. Meat Demand? American Journal of
Agricultural Economics 86(1):154–174.
Roberts, Tanya. 2007. WTP Estimates of the Societal Costs of
U.S. Food-borneIllness. American Journal of Agricultural Economics
89(5): 1183–1189.
Saghaian, Sayed H. 2007. Beef Safety Shocks and Dynamics of
Vertical PriceAdjustment: The Case of BSE Discovery in the U.S.
Beef Sector. Agribusiness: AnInternational Journal 23: 333–348.
Shogren, Jason F., John A. List, and Dermot J. Hayes. 2000.
Preference Learning inConsecutive Experimental Auctions. American
Journal of Agricultural Economics82(4): 1016–1021.
Stenger, Anne. 2000. Experimental Valuation of Food Safety
Application to SewageSludge. Food Policy 25: 211–219.
Thomsen, Michael R., and Andrew M. McKenzie. 2001. Market
Incentives for SafeFoods: An Examination of Shareholder Losses from
Meat and Poultry Recalls.American Journal of Agricultural Economics
82(3): 526–538.
Thomsen, Michael R., Rimma Shiptsova, and Sandra J. Hamm. 2006.
SalesResponses to Recalls for Listeria Monocytogenes: Evidence from
BrandedReady-to-Eat Meats. Review of Agricultural Economics 28(4):
482–494.
Appendix 1
Leading-brand chicken
According to Consumer Reports magazine, a recent study of
broilerchicken revealed that “campylobacter was present in 81
percent of thechickens, salmonella in 15 percent; and both bacteria
in 13 percent. . . .
Applied Economic Perspectives and Policy
382
-
Both salmonella and campylobacter can cause intestinal distress,
andcampylobacter can also lead to meningitis, arthritis, and
Guillain-Barrésyndrome, a neurological disorder . . . Among all
brands, 84 percent of thesalmonella and 67 percent of the
campylobacter organisms (tested)showed resistance to one or more
antibiotics. . . . The findings suggest thatsome people who are
sickened by chicken might need to try several antibi-otics before
finding one that works. . . . No major brand fared better
thanothers overall. Foster Farms, Pilgrim’s Pride, and Tyson
chickens werelower in salmonella incidence than Perdue, but they
were higher incampylobacter.”
Consumer Reports concludes that their tests reveal that if you
eat under-cooked chicken (less than 1658F) or have
cross-contamination to otherfoods from mishandling the chicken,
“you have a good chance of feelingmiserable.”
Ranger brand chicken
Consumer Reports magazine reports that “there was an exception
to thepoor showing of most premium chickens. As in our previous
tests, Ranger. . . was extremely clean . . . Of the ten samples
analyzed, 0 percent hadsalmonella and only 20 percent had
campylobacter.”
Do Consumer Responses to Media Food Safety Information Last?
383