GETTING THE MESSAGE: FRAMING FOOD RECALL
MESSAGES WITH PROSPECT THEORY TO
INCREASE CONSUMER PROTECTION
MOTIVATIONS
By
Gregory Paul Clare

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
RETAILING
2012

ABSTRACT

GETTING THE MESSAGE: FRAMING FOOD RECALL
MESSAGES WITH PROSPECT THEORY TO
INCREASE CONSUMER PROTECTION
MOTIVATIONS
By
Gregory Paul Clare
Previous research has shown that positive and/or negative message framing is a relevant
factor in influencing behavioral intentions. Recall message framing may be useful for increasing
responses to high risk food product recalls targeted to consumers. This study utilized a prospect
theory message framing manipulation to determine the impact of positive, negative, and control
condition messages on protection motivation theory constructs in an online experimental design.
The findings of the research indicate that loss framed messages demonstrate greater strength for
reducing maladaptive responses to recall messages. We also found that the paths for adaptive
responses to recall messages were different between the gain and loss conditions. Self-efficacy
influenced adaptive responses for gain framed messages while coping and locus of control
influenced adaptive responses for loss framed messages.

Copyright by
GREGORY PAUL CLARE
2012

ACKNOWLEDGEMENT

This dissertation is dedicated to Patricia Huddleston, my guidance committee chairperson.
Her patient support and direction have made the achievement of a doctoral degree possible on
many levels. When I started my journey in the academy she helped nurture my development as
an educator. In the ensuing years she worked to hone my skills as a researcher and to develop
the patient tenacity that is required for a successful career aimed at answering the big questions
that will hopefully benefit society through our research agenda. Dr. Huddleston, you have my
most sincere and grateful thanks for believing in me through both the lightest and darkest times
on our journey together.
In addition, I would like to thank my father for his tireless support of achieving a first in our
family, a graduate education culminating in a doctoral degree. Your strong belief in setting and
achieving goals has always been an inspiration to me and for this grounding spirit I am grateful.
Finally, I would be remiss if I did not recognize the other members of my guidance committee
who provided just the right paths at the right times to lead me on my journey. Dr. Calantone,
the consummate econometrician and statistician who helped assemble the nuts and bolts into a
functional machine. Dr. Rifon, the theoretician who teased out the many nuances of research
opportunities that the average human would surely miss along the way. Dr. Carlo, the partner
and friend whose keen eye was able to dissect and classify potential innovations in the research
agenda.

v

TABLE OF CONTENTS
LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES .............................................................................................................x
INTRODUCTION ...............................................................................................................1
CHAPTER 1
PROBLEM DEFINITION......................................................... ..........................................4
Dependent Variables.......................................... ......................................................8
CHAPTER 2
LITERATURE REVIEW ....................................................................................................9
Introduction ..............................................................................................................9
Theoretical Framework ..........................................................................................11
Prospect Theory .....................................................................................................11
Protection Motivation Theory ................................................................................20
Bridging Prospect and Protection Motivation Theories.........................................25
Hypotheses .............................................................................................................26
Message Framing and Threat Appraisal ................................................................26
Perceived Efficacy .................................................................................................31
Coping Appraisals ..................................................................................................34
Proactive and Reactive Coping ..............................................................................34
Protection Motivations ...........................................................................................37
Locus of Control ....................................................................................................39
Confidence in the message Communicator............................................................41
CHAPTER 3
METHODS
Study Overview .....................................................................................................43
Pre-test Methods ....................................................................................................43
Participants .............................................................................................................43
Design ....................................................................................................................44
Measures ................................................................................................................44
Data Collection Procedures....................................................................................45
Findings..................................................................................................................45
Main Experiment Methods ....................................................................................46
Participants ............................................................................................................46
Design ...................................................................................................................49
Instrument .............................................................................................................50
Threat Assessment ................................................................................................51
Perceived Efficacy ................................................................................................52
Locus of Control ...................................................................................................53
Protection Motivation ...........................................................................................53
Perceived Credibility ............................................................................................54
vi

Data Collection Procedures....................................................................................54

CHAPTER 4
ANALYSIS ........................................................................................................................56
Statistical Analysis .................................................................................................56
Demographic Data Analysis of Sample .................................................................56
Descriptive Statistics of the Survey Items .............................................................58
Measurement Model Characteristics......................................................................60
Pretest Analysis ......................................................................................................61
Main Study Analysis ..............................................................................................62
Confirmatory Factor Analysis................................................................................64
Manipulation Check ...............................................................................................65
Common Method Bias ...........................................................................................66
Formative Latent Construct Structural Test ...........................................................67
Hypothesis Testing.................................................................................................69
Path Analysis .........................................................................................................70
Hypothesis Results .................................................................................................71
CHAPTER 5
DISCUSSION AND CONCLUSION ...............................................................................77
Discussion ..............................................................................................................77
Maladaptive Protection Motivations Loss Condition ............................................79
Maladaptive Protection Motivations Gain Condition ............................................80
Maladaptive Protection Motivations Control Condition........................................80
Summary of Maladaptive Protection Motivations .................................................81
Adaptive Protection Motivations Loss Condition..................................................82
Adaptive Protection Motivations Gain Condition .................................................83
Adaptive Protection Motivations Control Condition .............................................84
Summary of Adaptive Protection Motivations ......................................................85
Implications............................................................................................................87
Recommendations for Future Research .................................................................92
Limitations .............................................................................................................95
APPENDICES ...................................................................................................................98
Appendix A Tables ...............................................................................................99
Appendix B Figures .............................................................................................135
BIBLIOGRAPHY ............................................................................................................150

vii

LIST OF TABLES

Table 1. Preliminary Survey ..............................................................................................99
Table 2. Survey Instruments ............................................................................................101
Table 3. Demographics ....................................................................................................109
Table 4. Descriptive Statistics..........................................................................................117
Table 5A. Manipulation Check Mean Comparison ........................................................120
Table 5B. Analysis of Variance Manipulation Check Combined Samples,
Control, Gain, and Loss Conditions .................................................................................120
Table 5C. Moderation Test of Predictor Variables Effects on Dependent Variables .....121
Table 6. Model Fit Statistics, EFA and CFA ...................................................................122
Table 7. Main Study Factor Loadings (EFA) .................................................................123
Table 8. Main Study Factor Loadings (CFA reduced model) ........................................124
Table 9. Item and Scale Reliabilities, Reduced Model (CFA) .......................................125
Table 10A. Manipulation Check Structural Formative Model for Theorized
Protection Motivation Dimensions (Control Condition) .................................................126
Table 10B. Manipulation Check Structural Formative Model for Theorized
Protection Motivation Dimensions (Gain Condition) ......................................................127
Table 10C. Manipulation Check Structural Formative Model for Theorized
Protection Motivation Dimensions Control Condition (Loss Condition) ........................128
Table 11A. Significant Multivariate Effects, Multivariate Analysis of Variance
(MANOVA) .....................................................................................................................129
Table 11B. Box’s Test of Equality of Covariance Matrices ...........................................129
Table 11C. Levene’s Test of Equality of Error Variances..............................................129

viii

Table 12. Hypothesis Test for Structural Model Path Analysis .....................................130
2

Table 13. R Summary by Experimental Condition ......................................................132
Table 14. MTurk Participant Experiment Comments Post-test ......................................133

ix

LIST OF FIGURES

Figure 1. Conceptual Diagram of Prospect Theory and Protection Motivation Theory
Experiment .......................................................................................................................135
Figure 2A. Informed Consent Message ...........................................................................136
Figure 2B. Invitation and Reminder Message .................................................................137
Figure 3A. FDA Press Release of Stimulus Cantaloupe Recall Message ......................138
Figure 3B. Control Group Message Stimulus .................................................................141
Figure 3C. Gain Message Stimulus ................................................................................141
Figure 3D. Loss Message Stimulus ................................................................................141
Figure 4. Graphic Stimuli (Control, Gain, Loss) Conditions..........................................142
Figure 5. Main Study CFA Model (reduced) ..................................................................143
Figure 6A. Manipulation Check Control Condition .......................................................144
Figure 6B. Manipulation Check Gain Condition ............................................................145
Figure 6C. Manipulation Check Loss Condition ............................................................146
Figure 7A. Structural Model Path Diagram Control Condition......................................147
Figure 7B. Structural Model Path Diagram Gain Condition ..........................................148
Figure 7C. Structural Model Path Diagram Loss Condition...........................................149

x

INTRODUCTION

Product recalls are pervasive in the consumer marketplace. However, consumers largely
ignore product recall messages due to a belief that illness will not impact them personally. These
beliefs are based on consumers' substantial food consumption experiences that occurred without
incident or injury. Likewise, many consumers believe in the reliability and safety of the food
supply, and the business entities that produce and sell food products. Sporadically, consumers
learn of defective food products sold by retailers and producers that are suddenly recalled. These
recall messages are sent from any of a number of communication sources for differing reasons
including a potentially severe health risk if the product is eaten, undeclared allergens, or
incidents of product tampering. Unfortunately, many consumers never receive recall messages,
and when exposed to recall messages they usually fail to react to the message to avert injury. As
a result, consumers are periodically and often unnecessarily exposed to injury due to the failure
of the recall communication process. Members of the food industry supply chain may benefit
from strategies to increase desirable responses to recall messages. Desirable responses are
defined as those actions by consumers that reduce the potential of injury. We are interested in
examining different methods for creating recall message content that improve consumer
responses which, if heeded, will reduce or eliminate the risks associated with defective food
products. Specifically, we want to determine the impact of recall message framing on increasing
a consumer's tendency to take protective actions and avert injury. In addition, our findings may
also help improve recall communications in other consumer product industries. Research on the
merit of recall messages that motivate consumer responsiveness to food product recalls has been
seriously neglected in the risk literature. To address this gap in the literature we will utilize

1

protection motivation and prospect theories to measure consumer protection motivation
perceptions and behavioral intentions when exposed to probabilistic gain and loss oriented
message content.
Both product recalls and product warning labels are a form of safety communication to
consumers. Since 1990, product recalls have grown over 80% per year, and as many as 27,000
deaths and immeasurable accidents are attributable to products later recalled for defects (CPSC,
2011). In addition, it is estimated that between 24 and 81 million cases of food borne diarrhea
incidents occur in the United States each year and cost between $5 and $17 billion (Wagner,
2011). Mead et al. (1999) estimate that food borne pathogens account for 76 million illnesses
and 5,000 deaths in the United States annually.

One can conclude from these food borne illness

statistics that retailers and manufacturers need to do a better job of communicating recalls and
eliciting appropriate responses to recall messages.
Deloitte Inc., conducted a survey of 1,110 US food consumers, and found that 57% of
Americans stopped eating certain foods, either temporarily or permanently due to food safety
concerns (Food Product Design, 2008). Likewise, the survey found that 73% of Americans
believed food recalls increased in the past year, and 76% were more concerned about food safety
than they were five years ago (Food Product Design, 2008). In spite of the aforementioned
statistics on food borne illness, consumers typically ignore food recall notices (Fox & Simao,
2009). After the Salmonella St. Paul outbreak of 2008 that involved contaminated tomatoes,
Rutger’s University Food Policy Institute conducted a telephone survey of 1101 consumers and
64% of respondents reported eating tomatoes in general during the recall timeframe and 38% of
respondents ate tomatoes included in the recall warnings (Hallman, Cuite, & Nucci, 2009). Of
those that ate the recalled tomatoes, 89% reported they were aware of the warning at the time

2

they ate tomatoes (i.e. recalled or otherwise) (Hallman, Cuite, & Nucci, 2009). These findings
raise the question: why do consumers ignore these warnings?
Public recalls of food are not only on the rise but appear to be more extensive than ever,
and it is certainly the public’s perception that the food safety oversight system seems to be
failing (Hemphill, 2009). Several surveys have shown decreased consumer confidence in the
food supply after product recalls (Food Processing, 2007; Haberkorn, 2007). Reasons for
ineffective recalls are that public notification is typically broadcast via the news media and
messages are delivered too late, often after the food has been purchased and eaten, often notices
are not crafted in a manner which influences the greatest number of consumers at risk of food
borne illness to take protective action.
In the aforementioned Rutger’s St. Paul Salmonella outbreak study, 90% of participants
reported that they received information about food through the popular press and television.
(Hallman, Cuite, & Nucci, 2009). Similar results were discovered in a 2005 study of U.K.
consumers related to food scares, their sources of information included: news media, government
notices, articles in newspapers and women’s magazines, television programs, and press releases
from food suppliers (Dawson, 2005). In Dawson’s qualitative research study, consumers
indicated that in the face of a more serious food safety concern, there is a greater tendency to
seek out scientific and government sources of information. Likewise, Dawson (2005) found that
the news media messages produced a dual effect of confusing consumers and undermining
confidence in the food supply. We will determine whether modifying the level of information in
a recall message increases consumer tendencies to take action and thereby reduce the potential
for injury.

3

CHAPTER 1
PROBLEM DEFINITION

The news media and popular press are important communication partners for retailers
trying to reduce injury from recalled products, and broadcasting recall messages is an important
public service. However, we believe that mass media recall notice should remain a secondary
means of communication. Once a retailer, manufacturer, or farm producer working in
partnership with government start a food recall notice; the supply chain partner closest to the
consumer's point of purchase should be the primary recall communicating partner. Despite
retailers closeness to the consumer at the end of the supply chain, retailer recall communications
typically lack important information to help consumers reduce risks of injury. Retailers and
manufacturers are interdependent partners in the supply chain and each provides channels of
information to consumers about products. Information from retailers may elicit higher
responses from consumers because of the closeness between the consumer and the point of
consumption (e.g. retailer). The transfer of products between consumers and retailers may
strengthen information exchanges through a consumer's perceived familiarity, trust, and loyalty
for the retailer or other psychological motivation states. On the other hand, the news media
offers consumers a one-way information transfer, and seldom allows the exchange of
information common in retailer consumer communications. The inability of news media
information to address specific consumer concerns or allow added information seeking underlies
the importance of this research. Developing a method for creating effective recall message
content which encourages consumers to take protective action if a food recall occurs is critical.
Behavioral intentions that lead to suitable responses to recalls should decrease financial loss for

4

retailers and manufacturers, increase the perceived importance of recall messages, improve
perceptions of food safety initiatives of businesses, and lessen of the risk of consumer injury
from food borne illness.
Many recall injury cases are preventable. Proper handling is defined as an action or
actions by the consumer to reduce the risk of food borne illness from consuming a recalled
product. Proper handling may involve: disposing of the product, returning the product to a retail
store, or following specific instructions about preparation (e.g. cooking the food at a specific
internal temperature), or other tactics. Proper handling implies a payoff to consumers in the form
of a gain (e.g. lowered risk of injury) or loss (e.g. greater risk of injury) and will depend on
personal choices as whether to engage in prescribed behaviors stated in a product recall message.
This study will provide hypothetical recall message content with the expected goal of stimulating
a consumer’s intent to properly handle recalled products. We will measure the effects of
message content on behavioral intentions and whether consumers perceive hypothetical recall
messages as sources of useful information for taking action and avoiding injury.
Most recall messages in the current marketplace provide necessary information about the
causes of the recall (e.g. Listeria, salmonella), but insufficient information about proper handling
that will reduce potential for injury. This problem is complicated by the fact that the
probabilities of injury differs among groups of consumers, as some segments possess unique
vulnerabilities to food borne illness which are related to age or compromised immunity.
The goal of manipulating recall message content using prospect theory is to assess the
impact of positive and negative message frames and their effect sizes on protection motivations.
Our goal is to identify message content that will have a positive influence on both vulnerable and

5

less vulnerable groups of consumers. Food industry supply chain members would benefit from
understanding which message treatments display larger effect sizes on protection motivations.
Recall communications are presumed to be a persuasive tool to decrease food borne
illness incidents but typically lack information about alternative strategies to minimize risks.
Product warning labels are ubiquitous on food products and offer detailed information to avert
injury such as 'use by' and 'best by' product expiration dates, 'refrigerate after opening' and
instructions for proper food handling such as preparation guidelines (e.g. internal cooking
temperature) and strategies to avoid spreading pathogens to other foods through cross
contamination. Warning labels are an attempt to reduce the risks from food borne illness through
proper handling.
Recall communications differ from warning labels because they are introduced into the
consumer marketplace selectively and may involve a single product, many products, or entire
classifications (e.g. cucumber, lettuce or tomato recalls). Recall communications are reactive
appeals to product failures with the goal of reducing injury. The primary goal of recall messages
is to influence consumers to identify and remove the potential risks from the food product by
throwing it away or returning it to a store. Recall messages communicated by mass media do not
typically suggest that through proper handling, the product may be safe to eat. The absence of
proper handling instructions or the focus on disposal or return of the defective product may result
in a perceived loss for consumers in time and effort. Messages offering information about how
to properly handle recalled products to avoid or cope with the potential loss or injury may be
perceived as potential gains by the consumer. Recall warnings typically advise of the risks from
the defective product in question, but offer little or no information to cope with the risks in a
proactive manner other than simply eliminating the risks by disposal of the product or returning

6

it to the store. Reinforcing the benefits of proper handling on reducing food borne illness in
recall messages may be perceived as a gain by consumers and increase their tendency to protect
themselves and their family.
Consumers are directly impacted by product recalls when they purchase a recalled
product. For example, a retailer may identify a defective food product, such as ground beef, that
may pose a hazard to a specific geographic area and try to notify consumers about the recall.
Retailers may also provide details of how to handle, return, or dispose of a recalled product in
the communication. Retailer approaches to notify consumers about recalls vary from low-tech
in-store signage to high-tech targeted communication encompassing specific consumer
purchases. Targeted communications are achieved by linking store loyalty program data to recall
notices. Recall messages sent within the various delivery channels may be received, processed,
or acted on by consumers. Even if the intended audience receives the recall message, they must
engage in a multistage process to avert risk. First, the consumer must interpret the message for
preventative action, store it in memory, recall it cognitively, and act on it in an satisfactorily to
avert risk.
Adding to the challenge of a consumer's multistage cognitive processing of recall
messages, retailers, producers and governmental agencies must deliver recall messages in a
technologically changing communication environment. Consumers are experiencing increased
daily time pressure and exposure to increased channels and numbers of communications each
day (e.g. television media outlets, Internet, e-mail, social networking, cellular text messages, and
cellular voice messages). Capturing consumer attention to a recall message because of these new
challenges to both message delivery and cognitive processing are theorized to be controllable, to
some degree, by manipulating recall message heuristics. Because of these reasons, theory driven

7

and empirically tested message content to increase consumer reactions to recall messages would
benefit businesses and consumers regardless of the recall communication channels they choose
to exploit.
Dependent Variables
Our goal is to understand what motivates consumers to respond to recall messages
(e.g. protection motivations/behavioral intentions) by manipulating recall message content. The
dependent variables include perceived credibility of the message communicator and protection
motivation. Perceived credibility of the message communication will measure the impact of the
gain/loss treatments coping responses and whether perceived effectiveness of the message
influences the protection motivations/behavioral intentions of the participants. Protection
motivation/behavioral intentions are based on a stimulus from a hypothetical recall message.
The hypothetical recall message exposures were presented to participants within two randomly
assigned experimental groups (i.e. each group receiving gain or loss oriented message content).
The proposed model is presented in Figure 1.

8

CHAPTER 2
LITERATURE REVIEW

Introduction
A study conducted by the Consumer Product Safety Council (CPSC) determined there was a
weak relationship between the threats of a potential recall hazard and recall effectiveness (CPSC,
1980). We are interested in finding out if message content has an impact on consumer responses
to a recall. Our goal is to test recall message content that motivates an suitable protection
motivation response from consumers. Discovering how to get consumers to pay attention to
recall messages and take appropriate action, by, increasing the motivations to protect themselves
is critical to improving recall responses.
Consumers are overwhelmed with messages daily, with estimates suggesting that in
1971, consumers faced 560 messages daily, increasing to over 3000 messages by 1997 (Shenk,
1997). New forms of communication such as the Internet and mobile technology are anticipated
to increase these message exposures for consumers. A consumer’s capacity to make sense of and
prioritize so many message exposures overwhelms their cognitive ability, and may be
characterized as information overload. Dealing with information overload requires consumers to
cut corners and apply heuristics within their cognitive framework to deal with this information
overload. Eagley and Chaiken (1984), created a framework for common heuristics employed by
consumers in decision making. Each of these heuristics is posited to influence consumer
cognitive processing of recall messages.

9

These top five heuristics include:
1. Trust in expert opinion (e.g. chief food scientist versus. activist; government entity
versus business); 2. Trust in the communicator’s perceived friendliness (e.g. personalized
greeting versus form letters); 3. Influence from the degree of persuasive arguments in a message
(e.g. coping strategy presented versus threat focus); 4. Influence by statistics employed in the
message (e.g. presenting statistics versus generalizations); 5. Influence from social cues and
message context (e.g. risk to self versus significant others and community (Eagley and Chaiken,
1984).
A consumer's exposure to and probable use of heuristic cues is likely to influence
perceptions of the usefulness of the message. In their study of persuasion to exercise, Jones et al.
(2003) used prospect theory and message frames to measure perceived source credibility by
manipulating credible and less credible sources and positive and negative information frames.
The researchers found that credible source and positive information influenced intentions to
exercise and that negative messages from a credible source influenced lowered intentions to
exercise. We will measure consumers' perceptions of the communicator credibility in the gain
and loss message conditions by testing the perceived trust and expertise implied in the message
from Eagley and Chaiken' (1984) common consumer heuristics and prospect theory. We
anticipate consumers' protection motivations were influenced by varying statistical probabilities
of illness in experimental message content. Likewise, we anticipate that by varying statistics in
the messages, consumers' perceived credibility of the message communicator will be influenced.

10

Theoretical Framework
Prospect Theory
Kahneman and Tversky (1979) conducted research in which participants were faced with
risky choices and found that people are more concerned about gains and losses during specific
events than protecting their net assets. According to prospect theory, people should respond
predictably to potential gains or losses, and prefer perceived gains to losses. People continue
from a cognitive reference point in their decision making and this reference point may be
described as an equilibrium state or status quo state. A person’s individual health status
perception (e.g. good health, bad health) is an example of a theoretical reference point. A Class
I recall message exposure will cause a consumer to create decision weights of the implied
reasons (e.g. efficacy, threat, coping) influencing whether they react to or ignore the message.
The consumer's cognitive weighting scheme influences perception of changes to a health
reference point, based on the losses or gains a consumer experiences from their choices.
Ignoring a recall message, eating a defective food product and afterwards becoming ill are an
example of a loss. This loss will cause a deviation from the health status reference point. On the
other hand, a consumer's reaction to a recall message by avoiding a defective food product and
remaining healthy is an example of a gain.
Prospect theory proposes that people tend to be risk averse with respect to gains and risk
acceptant for losses (Kahneman & Tversky, 1979). This finding is counterintuitive. People
prefer results which they are certain will occur. Losses are given more cognitive weight and
have more behavioral influence than gains. Adding to a person's typically counter-intuitive
behavior, people under-estimate low chance events. If the likelihood of death from a recalled

11

product is small, consumers are likely to tend not to react to a recall message. This tendency in
prospect theory inclines people to take risky choices.
Prospect theory suggests that people under weigh probabilistic outcomes compared to
outcomes stated with certainty by a communication source. This finding is referred to as the
certainty effect and influences risk aversion in choices of stated sure gains and risk seeking in
choices involving stated sure losses. According to the theory, individuals also display a tendency
to ignore information shared by all prospects presented in a choice scenario. Prospects are
defined as the possibility of achieving an outcome. The consumer presented with the same
information stated in different ways will apply cognitive weights and attribute meaning to this
information, finally making a choice based on their unique weighting scheme. This tendency is
referred to as the isolation effect and leads to inconsistent preferences when participants are
presented with the same choice communicated in different forms.
A consumer's decision making to choose a choice among risky choices is determined by
specific weights applied to those choices and idiosyncratic cognitive schema that define
perceived gains and losses. The consumer's weighting is not typically additive in nature and
greater weighted individual perceptions may more heavily influence choice than several causes
working together to reach an ideal decision (Tversky & Kahneman, 1992). This finding is
referred to as the reflection effect around the reference point and it may show some degree of
symmetry when testing opposing weighted choice statements between groups of consumers.
Symmetry assumes equality (e.g. good or bad health views) on opposite sides of the reference
point.

In other words, faced with a potentially disastrous health loss such as might be faced

when consuming a Class I recall food product, message framing content is likely to change the
consumer's reflection effect from risk aversion or risk seeking when presented with either a gain

12

oriented or loss oriented message treatment. This effect is dependent on perceptual variations
from the hypothetical health reference point caused by variations in message content offering
chances for gains or losses. The implied costs (e.g. information seeking costs) of reacting to the
recall message for loss oriented message content were higher than those in gain oriented
messages. If participants receive a loss oriented recall message, we expect them to display lower
wishes to react to the message (e.g. risk seeking) and averting injury. The contrary is expected
for exposure to a gain message (e.g. risk avoidance). Participants should show a greater want to
respond to the message to stay healthy.
Tversky and Kahneman (1981) explored the effects of an epidemic in which expert
opinion described a frame of people who would survive and a frame of people who would die. .
In their classic experiment the following choices were presented to participants:
1. 200 people will be saved
2. there is a one-third probability that 600 people will be saved, and a two-thirds probability
that no people will be saved
3. 400 people will die
4. there is a one-third probability that nobody will die, and a two-third probability that 600
people will die
Given these choices, 72% of participants chose alternative 1 as their preference, 28% chose
alternative 2. A second group of participants was presented with alternatives 3 and 4 and the
results were similar, 78% chose alternative 3 and 22% chose alternative 4. Conditions 1=3 and
2=4 in their experimental design. People clearly preferred the gain frame in this experiment that
presented lives saved as opposed to a tradeoff mentioning that people will die. However, more
people chose alternative four than three in the second condition because people preferred a risky

13

alternative with the opportunity to avoid losses versus less risky alternative three, a certainty of
loss. The participants were presented with one alternative in which 200 of 600 people’s lives
would be saved when impacted by an epidemic (survival frame). In the second mortality
alternative the participants were offered the alternative of a 33% chance of no people dying and a
66% chance that all 600 would die (death frame). In both cases probability suggests 200 lives
would be saved. Given these two choice frames, participants preferred the mortality frame as
their option, , lending credence to the preference for risk seeking behavior. This finding shows
the way in which people frame outcomes based on information presented to them, and reflects a
cognitive reference point analysis based on the information presented. For recall messages, the
consumer's choice to react involves processing and reacting to the stimuli presented. Providing
heuristic message cues in the gain message and removing or lessening them in the loss message
should increase or decrease consumer behavioral intentions to respond directly (protection
motivations).
Tversky and Kahneman (1992) changed prospect theory with cumulative prospect theory
with the findings from an experiment in which participants were presented with eight choices
(gains or losses) spaced between extreme outcomes (i.e. demonstrating certainty effect) such as
25% chance to win $150 and 75% chance to win $50. The researchers discovered that when
consumers evaluate risks with high probability of happening they are risk seeking for losses and
risk averse for gains. Conversely, they found that consumers faced with low probability events
are risk seeking for gains and risk averse for losses.
The approach of influencing consumers' perceptions of outcomes is similar to gambling
when gamblers and dealers weigh the potential probabilities of gains and losses as outcomes of
dice rolls with the goal of rolling two dies and achieving repeated outcome rolls of seven. The

14

chances of such an anticipated outcome are clearly defined by 36 possible rolls and 6
combinations of the dice that will produce a result of seven, or 16.7% of the time a seven was
rolled by the gambler. However, the gambler may apply various weighting biases on the
conditions under which the dice rolls will perform, such as perceptions of luck or dice rolling
skill which defy the rules of probability and dominate the choice to continue gambling. These
weighting biases are attributed to the gambler's reference point as potential gains or losses and
the outcome of rolls will move the reference point. A gambler is more concerned about losses
because they do not support his cognitive reference point and self-positivity bias. The gambler
will continue playing the game until self-positivity biases are disconfirmed through losses. The
dealer controlling the craps game may also employ perception management to influence the
gambler's weighting schema for rolling sevens by manipulating the gambler's perceptions of his
odds of rolling sevens which exceed probabilistic limitations. Heuristic fallacies such as
implying the table is winning or "hot" tonight, the roller is lucky compared to others, or the dice
are magical may influence perceptual weights of improved odds. In our study the "gambler" is
the consumer interpreting a recall message and deciding whether to react within a cognitive
weighting schema presented with factual information, but impacted by heuristic biases. The
recall message is similar to the dealer who is trying to influence the perception of the consumer,
that, by taking appropriate action they will gain lasting or better health. Positive message
content should favorably influence the consumer through gains to the health reference point, and
losses should produce the opposite effect.
The consumer's weighting attributions to probabilistic results are referred to as the selfpositivity bias and this bias depends on the theoretical health reference point's current state
(feeling good or bad). The self-positivity bias suggests consumers are less likely to ignore recall

15

messages if their cognitive health reference point is perceived as bad health. Consumers who
sense that their health reference point is good are more likely to discount a recall message when
it is framed negatively, but react when it is framed positively.
The health risks from Class I recall products are great if the defective food product is
eaten by any consumer, and probabilistically higher if consumers are members of a risk group.
Despite the fact the risks of food borne illness are well-known, many consumers gamble with
their health by ignoring the information presented in recall messages. For food recalls, similar
discounting effects such as the consumer's view that they are invulnerable to disease or were
lucky enough to avoid food borne illness because their cognitive reference point for their health
status is good are critical fallacies. Consumer perceptions of losses or gains involve four
dimensions we plan to address in our experimental design:
1. Reference point dependence: Consumer's view food borne illness resulting
from a recalled product with a change in the status quo of their normal food
consumption. In other words, food borne illness is a low chance event for most
consumers, and they will evaluate message content based on their prior experience
with food borne illness.
2. Gain and loss satisfaction: Consumers frequent exposure to recall messages
presented by the news media imply the law of decreasing returns postulated in
prospect theory for both good and bad responses to recall risks, but may be
influenced by targeted message design.
3. Loss aversion: Losses such as becoming ill from a food recall are weighted
more than the behaviors for averting a food borne illness.

16

4. Risk aversion and risk seeking: Consumers presented with recall messages
signaling gains from behaviors to avoid illness were risk averse and when
presented with losses were risk seeking. Proper handling stimuli will likely cause
the consumer to take preventative actions (risk aversion). Less useful stimuli (i.e.
key heuristic cues missing) will likely cause consumers to ignore the risks of food
borne illness, and take little or no actions to mitigate the risks or perhaps they
might even consume the recalled product (risk seeking).
Rothman and Salovey (1997) suggest that positive framing is more effective in
motivating preventive actions (e.g. sunscreen use) and loss framing is more effective for
encouraging detection actions (e.g. self-breast examination). Recall messages may be described
as preventive action communications. Donovan and Jalleh's (2000) child immunization study
found that framing has a significant effect on attitudes when framed positively versus negatively,
with greater acceptance of the message when framed positively.
A recent study showed that consumers’ reaction to fear appeal messages increased when
the consumer’s perceived utility (usefulness) of the message content was high (KnoblochWesterwick et al., 2005). Usefulness is a subjective perception of value or gain in knowledge.
Useful information may include message cues that try to aid consumers in avoiding risks and
preventing injury. Useful information may also be perceived as a gain to customers by helping
them avert risks in objective ways, or is less influenced by emotions and personal biases. Useful
gain oriented information in recall messages will likely influence coping appraisals in individuals
by helping them avoid potential risks. To test the power of this assumption, both useful gain and
less useful loss message cues were included in the design.

17

An example of gain information from a Food Safety and Information Service (FSIS)
press release is the following:
Consumers should only eat ground beef or ground beef patties that have been
cooked to a safe internal temperature of 160° F, whether prepared from fresh or
frozen raw meat products (FSIS, 2011).
The following is an example of less useful loss oriented information:
The only way to be sure ground beef is cooked to a high enough temperature to
kill harmful bacteria is to use a thermometer to measure the internal temperature.
FSIS messages offer greater detail about the implications of a recall to consumers than the
typical news media recall message. The press release offers information about proper handling
of the at risk product to avoid injury while most news media sources will tend to stress some
variation of the following information:
A total of five confirmed Shiga-toxin producing E. coli O157: NM
cases and four probable cases have been reported in Lapeer,
Genesee, Isabella, and Sanilac counties. Illness onset dates range
from July 18-30. Those affected range in age from 15-88. Each
clear plastic bag and retail package bears the establishment number
“EST. 33971” within the USDA mark of inspection. The products
subject to recall were produced on July 7, July 15, July 21, July 28
and Aug. 4, 2011, and sold to retail establishments and restaurants
in Armada, Lapeer and North Branch, Mich. The products were
also sold directly to consumers from a retail establishment owned
by McNees Meats and Wholesale (Food Poisoning Law Blog,
2011)

The above information only provides a starting point for consumers to fittingly respond to the
health risks from the ground beef subject to the recall. While news media sources may provide
some information for suitable coping appraisals by communicating proper handling information,
these messages tend to be inadequate to increase protection motivations. This is because of the
main focus on the- who, what, when, and where of the recall and not enough information about

18

HOW to react to avoid injury. News media sources seldom offer extra information about ways to
cope with potential injury such as proper handling information.
News media sources condense recall information and often fail to provide necessary
information to help consumers avert injury, but retailers typically offer consumers even less
information than government entities or other members of the supply chain, and may be
contributing to injuries that could be avoided. For example, in Figure 3a a recent FDA recall
notice is provided which offers high levels of detail about the severity of the risks, vulnerability
to the risks, and coping strategies to deal with the risk. Contrast this information with the
following recall notice on Kroger.com, a large national grocery store chain:
KROGER TURKEY GRIND 85/15 CASE READY, 48 OZ
Affected in Georgia, South Carolina, Alabama and Knoxville, Tennessee,
Indiana (except SW IN, -Evansville-), Illinois, Eastern Missouri, Central
and Northwest Ohio and N. West Virginia panhandle, Greater Cincinnati,
including Northern Kentucky, Dayton, OH and South Eastern Indiana,
Greater Memphis, TN, Arkansas, Mississippi and Western Kentucky,
Michigan, Greater Louisville (including Indiana), Lexington, and
Nashville, TN, North Carolina, Virginia, Eastern WV, Eastern Kentucky,
SE Ohio, Texas and Louisiana; Dillon’s, Food 4 Less, Fry's, Jay C, King
Supers and Ralphs stores.
Reason: These products may be contaminated with Salmonella and, if
eaten, could result in severe illness to those individuals who may consume
this product (Kroger, 2012).
This recall notice provides the possibility of injury if eaten and the severity of the risks, but
provides little no information to consumers about how to address the risks. An example of
useful information added to Kroger’s recall notice includes ways to deal with the problem, such
as, proper cooking temperature, throwing out the product in a sealed container, or returning it to
the store for a full refund.

19

A consumer's motivation for an appropriate response to a recall message stems from an
individual’s perception that reacting to the product recall will reduce the threat of injury. We
theorize the content of a recall message will similarly lessen the perceived threat of injury. The
stimuli recall message content for this study will include Class I recalls, which potentially pose a
significant risk to consumer health if the defective product is consumed.
Prior research has suggested that message-based judgments are influenced by the
representativeness heuristic and not the availability bias (Tversky and Kahneman, 1974). The
representativeness heuristic suggests that consumers access memory for a comparable event in
their experience and assume the probability of occurrence is similar.

The information

presented in a recall message must be representative of the consumer's available memory of
similar events. If the information in the recall message is varied in gain and loss conditions we
theorize the consumer will adjust for the differences by using their availability bias. The
availability bias implies that to determine the likelihood of events individuals make a judgment
based on imperfect memory. Potential events such as serious illness from recalled food products
may result in denial (e.g. self-positivity bias) and therefore the thoughts are less available to the
consumer.

Protection Motivation Theory
Protection motivation theory (Rogers, 1983) is a widely tested framework for understanding
health communication strategies. PMT examines the effects of persuasive communications,
message sources, and message credibility on consumer cognitions which mediate behavioral
changes. These consumer behavioral changes are posited to be desirable (adaptive) or
undesirable (maladaptive) in nature.

Substantial research utilizing PMT has been conducted to

20

gain an understanding of consumer motivations for averting health risks; these studies have
employed various manipulations of health oriented messages (Rogers, 1983; Tanner, Hunt &
Eppright, 1991; Rogers & Prentice-Dunn, 1997; Witte & Allen, 2000). PMT has been used to
explain consumers’ motivations to take protective actions during fear appeals and their coping
strategies to deal with risk events. The ability of information to impact a consumer's beliefs
determines the effectiveness of health oriented information.
The first stimulus in PMT is information offered to consumers about a specific risk.
Eliciting an adaptive response to a health threat requires four elements: perceived severity of the
threat, vulnerability to the threat, response efficacy or the power of the message to create a
response to the threat, and self-efficacy or the willingness and ability of the individual to react to
the message (Bandura, 1977). Fear appeals which produce changes in behavior have
demonstrated mixed results in the PMT research stream. The factors which influence initial
adaptive behavioral changes which revert to maladaptive behaviors are not clearly understood.
Rogers (1983) suggested that when the costs of ongoing behavior change exceed the rewards of
performing the behavior, consumers simply stop performing the behavior after some period of
time. The valence of the threat message severity is suspected to have an impact on ongoing
behavior change. There is some support in the literature for this claim when consumers function
within various stages of contemplation of possible threats and their effects (Keller & Block,
1998).
For example, the transtheoretical model proposed by Prochaska (1994) combines stages
theory and PMT proposing that consumers move through six stages of behavior change, often in
a non-linear fashion, with transitions back and forth between stages based on feedback between
perceptions of costs and benefits of ongoing behavioral change. These stages include: pre-

21

contemplation, contemplation, preparation, action, maintenance, and termination. Each stage
measures the consumer's readiness to act when faced with a message stimuli aimed at behavioral
change. Researchers have demonstrated that many cognitive factors may effect these stages, such
as threat severity, susceptibility, and efficacy perceptions. Keller and Block (1998)
demonstrated that consumers presented with stimuli suggesting condom use for "safer sex" to
avoid STDs and are predisposed to never change their behaviors (e.g. precontemplators) because
of a strong disregard for change, have a greater tendency to take protective actions when exposed
to higher gain threat vulnerability messages compared to loss threat vulnerability messages. This
finding suggests that one effective way to change behaviors of consumers who are resistant to
change is the increase communications of their vulnerability in message oriented stimuli.
Interestingly, the researchers did not demonstrate the same behavior modification tendencies by
increasing severity or efficacy message valence for precontemplators (Keller & Block, 1998).
According the transtheoretical model, contemplation stage participants are those who
understand that adopting maladaptive behavioral responses will likely lead to future undesirable
states and accept the need to adopt adaptive behaviors, but these consumers are not ready to
change their behavior in the near future. For this group, Keller & Block (1996) demonstrated
that by manipulating the severity of the consequences portion of a message related to condom
use produced greater behavioral responses for participants in this stage. Communicating a fear
oriented stimulus to consumers potentially at risk of food borne illness will either cause
consumers to evaluate and react to the message by coping or to completely ignore the message
(Keller and Block, 1996; Rogers, 1983; Tanner, et al., 1991; Witte, 1994). We need to better
understand the determinants of perceived ongoing behavior change by carefully measuring the
impact of threat message content on consumers. We need to better understand the following:

22

1. What is the consumer's initial response to recall messages?
2. What is the anticipated continued response to recall messages?
3. What are the perceived costs of responding to recall messages?
4. What are the declining benefits of responding to recall messages?
5. How does recall message content influence habitual behavior?

To understand these factors it is necessary to measure several key aspects of consumer behavior
as it relates to recall message responses including: consumer personality traits, consumer life
situations provided by demographic indicators, message content strategies to move consumers to
a prescribed "desirable" state of responding to recall messages as opposed to the "undesirable"
state of avoiding them, the innate optimism of the consumer to enact behavior change, and
perceived self-esteem that they are in control of their lives versus at the mercy of external
factors.
Using PMT, Bender et al. (2006) utilized a transtheoretical model to assess ways that
consumers might lessen the risk of wildfire damage to their homes. The researchers suggested
actionable strategies such as spraying a roof with water during fire events and trimming excess
kindling brush near homes. Consumers were influenced by different dimensions of PMT based
on their cognitive stage. For example, pre-contemplators were most influenced by vulnerability
and contemplators by severity. Action stage participants had greater involvement with the risks
and also threat assessment (vulnerability and severity), were influenced by self and response
efficacy. Our purpose in this study is to begin to understand the impact of message phrasing on
threat and coping assessments and whether these phrasing approaches are perceived as more
credible sources of information to influence protection motivations.

23

Studies have shown the message framing and an individual’s cognitive appraisals do not
have the same impact on persuasion (Milne, Sheeran & Orbell, 2002). In one study, the authors
measured the impact of specific, as opposed to general, message treatments to increase
participant perceived relevance of regular exercise to their personal lives (Milne, Sheeran &
Orbell, 2002). They found that while PMT variables demonstrated a significant effect on
behavioral intentions, they did not impact subsequent exercise behavior. This phenomenon is
referred to as a discontinuation of expectations.
When attempting to perform continuing behavioral change routine performance of
behaviors, like responding to recall messages, will produce favorable outcomes that meet or
surpass the consumer's consumption oriented expectations of not getting sick when eating food
products they purchase. On the other hand, avoiding purchases of certain classifications of
recalled product (e.g. tomatoes) may not produce feelings of satisfaction to the same degree as
reacting to a specific recall message offered to consumers at a given time. Consumers must
resolve the lasting behaviors of avoiding a classification of food they would like to buy, but don't
because they are uncertain the product (s) are safe to eat. Barring a recall notice to confirm the
consumer's expectations of risks, and, routinized behavior of avoiding consumption of the
classification, we hypothesize that consumers will display a discontinuation of expectations to
continue their behaviors of avoiding products. This is because of the tension between the
consumer's wish to consume a specific classification of food, and placing limits to that
consumption caused by increased threats presented in the recall message. We do not understand
the temporal value of a recall message and how it impacts ongoing behavioral change. Will
avoiding certain food products be impacted by a single recall message, and if so, for what time
period? This avoidance effect occurs for some unknown period of time for consumers who may

24

avoid a classification of recalled products because of prior recall events, or consumers might
suspend their expectations of the risks, and return to consuming the products. In the event the
product is recalled, they may again engage in preventive behaviors for a time and then resume
consumption. We hope to influence the consumer's feelings of satisfaction by reacting to
specifically framed recall messages and stimulate ongoing performance of behaviors designed to
lessen injury from food borne illness. Because of these challenges, positive outcomes are
suspected to display less influence on behavior change than targeted elaboration of threat
susceptibility or severity in recall message communications. We plan to examine behavioral
intentions to find out if gain and loss message framing impacts participants' cognitive appraisals.
We believe the content of the message will have a significant influence on a consumer's
cognitive appraisals when controlling the message framing approach.

Bridging Prospect Theory and Protection Motivation Theory
Threat appraisal is defined as the degree to which one realizes the risk as a threat to
oneself (Witte, 1994). Threat appraisal is probabilistic, in which individuals stimulated by a
message evaluate the odds of injury and are motivated to choose protective or maladaptive
behaviors. Guided by prospect theory, changing the inherent probabilities of injury by
manipulating the message content with gain and loss frames, should produce predictable threat
appraisals. These threat appraisals should result in risk seeking (maladaptive) or risk aversion
(protection) in individuals. Both prospect theory and protection motivation theory propose a
probabilistic evaluation of risks and attempt to measure an expected result such as choosing a
protective behavior. However, no research has tried to measure the desired symmetrical
responses of individuals when faced with both a gain oriented and loss oriented message; nor has

25

the effect size of gain and loss oriented messages been evaluated within a PMT framework to
discover the impact of message framing on behavioral intentions and protection motivations.
Bridging both theories for a deeper understanding of consumer persuasion, using health
messages, is a critical goal of this research. If the expected reflexive symmetry suggested by
opposite effects (e.g. gain or loss) on protection motivations is discovered, then recall message
design approaches should improve recall response rates. The theoretical health reference point
posited from prospect theory should influence adaptive responses, and this finding will offer
food industry firms a theoretically useful best practice for developing recall message content that
promotes adaptive responses and minimizes maladaptive responses. We believe that recall
content that offers consumers useful information such as proper handling advice presented by
experts using statistics with convincing arguments and social impact cues, were more effective
than traditional news media recall notices that typically lack these cues.

Hypotheses
Message Framing and Threat Appraisal
PMT divides a person's threat appraisals into two dimensions: perceived susceptibility and
perceived severity of a named threat that is either supposed or tangible in nature. People possess
unique vulnerability to risk in life and certain causes such as age or immunity may increase their
vulnerabilities to threats. How people cognitively process threats and determine the probability
of being impacted by the threat may be mediated by their unique vulnerabilities which we can
determine through comparing demographic self-reports and measuring the differences, if any,
from a sample of consumers. Threats may be perceived as more obvious to some people than
others and show different degrees of influence on protective behaviors. For example, an elderly

26

married couple may perceive different degrees of susceptibility to disease and the severity of
consequences of that disease than a college student because of their ages and/or different life
experiences and genetic predispositions to illness. Rogers (1983) suggests a positive linear
relationship between a consumer's choice of protective behaviors influenced by a person's
parallel cognitive processing of perceived severity and susceptibility and efficacy factors (i.e.
self and response). Keller and Block (1998) studied condom use influence on protective
behaviors to avoid STDs and determined that when participants had higher perceptions of
vulnerability to illness elicited by message content, they exhibited greater intentions to use
condoms than did manipulations of greater severity, self, and response efficacy. This finding
complicates the proposed additive model of PMT by suggesting that consumers are in different
cognitive states of readiness to take action when presented with preventative means to change
behaviors. For this reason, we are interested in assessing the impact of threat and coping
appraisal variables combined with exposures to gain/loss message on protection motivations.
Susceptibility or vulnerability is defined as the perceived likelihood that consumers eating
a recalled product will contract food borne illness and thereby negatively impact their health
reference point. Threat susceptibility perceptions are likely to be strongly influenced by the
consumer's prior experiences with a specific threat (Dorn & Brown, 2003). If food borne illness
is not a typical experience for a consumer s/he is likely to mistakenly assume they are less
vulnerable to food borne illness. Many Class 1 recalls will cause illness regardless of whether the
individual perceives they are less likely to be injured. If a person encounters a recall message for
which s/he perceives low susceptibility to the risks then s/he were less likely to take preventative
actions to avert the risks than individuals who perceive high susceptibility to the risks of food
borne illness.

27

Severity is defined as the perceived probability that severe illness will occur from
consuming the recalled product. Consumers perceive threats within the framework of severity
and susceptibility by determining some amount of risk they are willing to accept in their lives.
For example, drivers may not wear seatbelts for any number of reasons on various occasions.
During these occasions, consumers cognitively create lower anticipated probabilities that they
will be in a car accident (e.g. susceptibility) and how badly they will be injured from not wearing
their seatbelt if an accident occurs (e.g. severity). The two perceptual states (i.e. severity &
susceptibility) are important interdependent considerations when consumers assess risks.
However, the nature and specific weighting scheme that consumers apply to myriad threat
situations they encounter routinely is not clearly understood, nor is it understood as it relates to
food recalls. What is clear from prior research is that consumers have a tendency to accept
greater risks when they perceive severity and susceptibility to be lower (Dorn & Brown, 2003).
Consumer reactions to specific threats, such as food borne illness identified in recall messages,
that are presented with greater severity and susceptibility to injury are more likely to influence
behaviors than those with lower severity and susceptibility to injury (Pyszczynski et al., 1997).
Our study will utilize Class I recall stimuli which will typically cause severe illness to
large proportions of the population if the recalled product is eaten. In other words, the severity
of the hypothetical recall risks to participants will be controlled by communicating risks to health
in the message content. The message framing effect has been examined using both protection
motivation and prospect theories with varied results. Donovan & Jalleh’s (2000) study of
parent’s perceptions of a hypothetical new vaccine found that positively framed messages
produced a greater impact on behavioral intentions than negatively framed messages. This
finding is inconsistent with prospect theory suppositions that negatively framed messages were

28

more effective than positively framed messages. Researchers have demonstrated the general
effectiveness of positively framed messages on consumer behavioral intentions pertaining to skin
cancer prevention and child safety seats (Marteu, 1989; McNeil et. al., 1982; Rothman et. al.,
1999). Other researchers have produced the opposite effect, with negative messages more
favorably influencing behavioral intentions on smoking cessation, exercise, breast examination,
and mammogram screening. (Wilson et. al., 1990; Robberson & Rogers, 1988; Meyerowitz &
Chaiken, 1987; Banks et al., 1995). These findings have been attributed to different consumer
perceptions of risk detection and prevention behaviors (Donovan & Jalleh, 1999).
PMT research has examined preventative behaviors that may impact the individual or
others (e.g. an individual smoking versus exposure to second hand smoke). Class I recalls are
not solely an individual concern since others, such as family members or friends, may also eat
the defective product

According to PMT, consumers will process the information about a Class

I recall cognitively and will appraise the threats based on the information provided in the
message influenced by a perceived threat severity and susceptibility. Consistent with prior
framing research, we expected loss frames to demonstrate a greater influence on perceived
susceptibility and severity than gain frames (Wilson et. al., 1990; Robberson & Rogers, 1988;
Meyerowitz & Chaiken, 1987; Banks et al., 1995). We also propose that since recall messages
are read by consumers (detected) prior to preventing the risk of illness, loss oriented messages
will be more effective for achieving adaptive responses than gain oriented messages.
Food recalls appear to demonstrate discounting effects of perceived risk severity based on
the large numbers of consumers who continued to eat tomatoes during the Salmonella St. Paul
outbreak (Hallman, Cuite, and Nucci, 2009). Loss oriented message framing information
provided to participants is suspected to worsen this effect. Consumers are likely to discount

29

recall messages because of their substantial experience of consuming food without injury. This
reaction is because of the fallacy that past events predict future outcomes. When exposed to loss
oriented messages, we expect participants will weigh susceptibility and severity of potential food
borne illness due to cues in the message. Similarly, gain oriented protection information is
suspected to increase the consumer’s tendency to discount the recall message. Prior research has
shown that: expert opinion, friendliness, persuasive arguments, statistics, and social cues are
useful in increasing consumer responses to recalls (Eagley & Chaiken, 1984). It is interesting to
note that 70% of consumers in Hallman, Cuite, and Nucci (2009) study received information
about the tomato salmonella outbreak from television with one-way message delivery and
limited framing information and 64% of respondents reported eating tomatoes during the recall
time frame. There may be parallels between the method of delivery and content of recall
message content and the inclination of consumers to discount the risks to health from the recalled
product.
Menon, Block, & Ramanathan (2002) showed the self-positivity bias (e.g. invulnerability
to illness) limits a consumer’s capacity to process messages. These researchers presented
participants with information suggesting that a disease was easy or hard to contract (e.g.
perceived susceptibility) and showed that consumers assumed that a disease is unlikely to affect
them. Likewise, as the frequency of messages increases, consumers were likely to begin to
discount the prior assumptions of invulnerability and begin to adopt protective behaviors (Menon
et. al. (2002). Participants, primed with a loss oriented message should produce a greater threat
perception than those exposed to gain oriented message.
Without a message frame, such as typical recall messages the impact of self- positivity
bias and perceived invulnerability are expected to be worsened. Therefore, when subjects read a

30

typical recall message, they are more likely to assume that foodborne illness will not impact
them, and ignore the message.
H1: Susceptibility inversely influences maladaptive protection motivations in the control
condition.
H2: Severity inversely influences maladaptive protection motivation in the control condition.

Perceived Efficacy
Perceived efficacy measures the ability of the individual to respond to the recall message
and his or her ability to control the implied risks in the recall message. Perceived efficacy
contains two component dimensions: response efficacy, which is a belief that some action on the
part of the consumer will avert the risk and self-efficacy which involves the consumer’s
willingness and ability to respond (Bandura, 2000). Response efficacy and self-efficacy are
probabilistic perceptions of the consumer's response costs which mediate coping with the threat,
and both dimensions contribute to perceived efficacy (Block & Keller, 1995). Response cost
perceptions may include implied and actual costs of taking action to avert a risk and the
consumer's assumption that accepting these response costs will produce a favorable result (e.g.
lessened risk of injury). According to PMT, when weighing response costs for reacting to a recall
message several dimensions are perceived by the individual. These include:
1. The individual must believe they are personally able (e.g. impacted, exposed) to react to the
threats in the recall message (self-efficacy). (Bandura, 1999)
2. The individual must be personally motivated (e.g. empowered, willing) to react to the threats
in the recall message (self-efficacy). (Bandura, 1999)

31

3. The individual's response (e.g. searching, returning, throwing out, additional information
seeking costs) to the threats in the recall message will produce the desired outcome of reducing
or eliminating exposure to illness (response efficacy) (Keller and Block, 1995).
We theorize that gain oriented information will have a positive impact on participant's
response perceptions related to self-efficacy and favorably influence reactions the risks. Gain
oriented information will improve consumer perceptions of self-efficacy and response efficacy.
This relationship is suspected because self-efficacy is a perception of how effectively consumers
can handle the risk situation, but these perceptions are a forerunner to effective responses.
Therefore, according to prospect theory, consumers will prioritize perceived gains over losses as
the way the implied costs of the willingness to respond. Self-efficacy differs from locus of
control because consumers believe internal influences like their personality, choices and thoughts
are controlling their behavior, as opposed to their perceived ability to act. Bandura (1977) found
that when participants assess the chances of an event occurring, their success depends on
perceived self-efficacy. In addition, Bandura determined there is a positive relationship between
self-efficacy and what he terms the outcome expectancy which we refer to as protection
motivations. Beck and Lund (1981) found that self-efficacy was a significant predictor of
participants' intentions to adopt preventative dental behaviors. In addition, the authors found that
the threat of gum disease in addition to participants' self-efficacy to perform the preventative
coping appraisals were significant predictors of protection motivations. This finding mirrors
other research by Roger, Deckner & Mewborn (1978) suggesting that participants require selfefficacy and some degree of response efficacy to achieve a coping appraisal and eventual
protection motivation.

32

Ignoring a health message is theorized to have a direct relationship with the selfpositivity bias and a consumer's belief that injury will not impact them personally (Patel et al,
2006). We posit that if the threat is not perceived as adequately severe, compared to the
response costs of reacting to a product recall, then the consumer is less likely to respond to a
recall message (e.g. maladaptive coping response).
Self-efficacy and response efficacy are linked to the coping appraisal and are probable
mediators of behavioral intentions to cope with risks (Maddux & Rogers, 1982). Block and
Keller (1995), in a two part study, examined the impact of stimuli that aroused and persuaded
participants to cope with the risks of human papilloma virus (HPV). The stimuli were framed in
positive and negative message treatments and low/high response efficacy. In their first study, the
researchers found that low response efficacy/negative frames were more persuasive than positive
frames due to a manipulation that preventive actions such as condom use would not reduce risks
to health However, in high response efficacy conditions positive and negative frames were
equally persuasive (Block and Keller, 1995). In their second study they attempted to replicate the
results from the first study using skin cancer detection techniques as the protection motivation.
Contrary to study one, the researchers found higher efficacy influenced behavioral intentions
more strongly than low efficacy.

The researchers account for these differences by suggesting a

difference between participants cognitive appraisal of health messages which are aimed at
preventative procedures to avert an illness as opposed to detection procedures. This finding
posits prevention messages are associated with high efficacy and detection messages are
associated with low efficacy. Since recall messages are detected by consumers once they are
read, they are necessary precursors to prevention behaviors. Recall messages may be assumed
for this reason to be low-efficacy and we anticipate similar findings to Block and Keller (1995).

33

H3: Self-efficacy inversely influences maladaptive protection motivations.
H4: Self-efficacy influences adaptive protection motivations in the gain condition.
H5: Response efficacy inversely influences maladaptive protection motivations.
H6: Response efficacy influences adaptive protection motivations.

Coping Appraisals
Milne, Sheeran and Orbell (2000), completed a meta-analysis of research studies utilizing
PMT to evaluate the usefulness of the theory in explaining behavior modification stemming from
behavioral intentions. The findings, involving 27 studies and 7,694 participants, demonstrated
that individuals weighed the costs of responding (i.e. coping appraisal) most heavily as part of
their decision making criteria. Self-efficacy (willingness and ability to respond) was the second
largest factor, response efficacy (reaction to the message produces a desirable outcome) was
third and perceptions of their vulnerability (threat assessment) was fourth in influencing coping
appraisal.

Proactive and Reactive Coping
Reactive coping is defined as consumer’s effort to deal with stressful events. These events may
be influenced by prior experience or present circumstances. For example, consumers may have
previously experienced an incident of foodborne illness. When faced with a food recall message,
the person has the choice of accepting the risk and avoiding action, or taking action to avoid
harm. Coping relies on self-efficacy because the person must feel that they are willing and able
to react, before making a decision to react to a risk situation.

34

Proactive coping is distinct from reactive coping because it involves developing a plan of
action to avoid risks before they happen. People who proactively cope with risks, such as
preparing emergency kits to cope with natural catastrophes are likely goal oriented and view skill
development as a challenge. A challenge for persons who exhibit proactive coping is to
understand the degree of susceptibility to risks and the level of severity of those risks to them.
Until a person perceives a threat with potentially negative impact on their health reference point,
they are likely to ignore the risks. However, as the number of threat messages increases, people
will be less likely to discount the messages and more likely to adopt a plan to avoid injury
through proactive coping strategy. This supposition mirrors Menon et al. (2002) finding that
consumer’s prior assumptions of invulnerability will decrease as they perceive higher
probabilities of risks impacting them personally.
For perceived efficacy to be effective at influencing the coping appraisal, consumers
must believe that performing personal actions will reduce the risk of injury. We posit that
participants must perceive recall message content as useful and actionable to avert injury. In
addition, reacting to a fear appeal may be hindered by the framework of message delivery. For
example, the Surgeon General’s warning on packages of cigarettes may not be evaluated and
reacted to by consumers despite the clear linkage between the prominent warning message on the
packaging and health consequences. This effect is likely due to the self-positivity bias and
generality of the delivery framework as opposed to a carefully crafted and directed message
incorporating gain and loss cues (Eagley & Chaiken, 1984).

Rogers and Mewborn (1976) found that fear appeals in communications were unrelated
to intentions to adopt the recommended action. Recalls may be classified as fear appeals because
of the message content, these messages typically emphasize anticipated outcomes from the risk,
35

but provide insufficient information to properly cope with the risks. The effect of emphasizing
probable outcomes are likely to elicit some degree of fear in subjects. We plan to test whether
gain or loss stimuli that magnifies anticipated outcomes for subjects influences the perceptions of
the risks involved in the recall message. Likewise, we will test whether including coping
strategies to avert food borne illness (i.e. throw out or return to store for refund), will influence
perceptions of the risks. We believe that favorable coping appraisals are more likely to be
elicited from gain oriented messages than loss oriented messages because consumers will
perceive higher value of prevention information (i.e. gains) as opposed to detection information
(i.e. losses) . Rogers and Mewborn (1976) also found that if recommendations were presented as
a means to effectively avoid a risk, individuals are more likely to adopt the recommended
behaviors. In addition, they found that perceived ineffective recommendations to avoid risk,
involving higher probability threats, decreased behavioral intentions to act on the messages
(Rogers & Mewborn, 1976).
Providing consumers with information to help them cope with risks is a vital component
of behavioral change. Protection motivation theory has demonstrated this relationship. A study
by Neuwirth, Dunwoody, & Griffin (2000) explored fluorescent lighting hazards under
conditions of high/low threat likelihood, susceptibility, and response efficacy conditions and
their impact on protection motivations. The coping strategy was to wear sunglasses to mitigate
the health risks. The researchers found a significant relationship between the threat
communication, response efficacy, coping appraisal, and protection motivations in the high
threat conditions. We anticipate similar findings for loss framed (high treat) messages influence
on coping when proper handling stimuli is provided in recall messages. Rogers and Mewborn

36

(1976) demonstrated that as the perceived effectiveness of the proposed coping action increased,
participants were more likely to adopt safe driving (preventative) recommendations.
Coping appraisals are predicated on the consumer's willingness and ability to respond to
a recall risk, along with a belief that such a response will reduce food borne illness potential if
the prescribed behaviors (i.e. proper handling) are followed. In conditions where self-efficacy
perceptions are high, we anticipate that consumers will more favorably perceive useful gain
oriented coping strategies in the message content. Useful gain information should positively
influence coping appraisals as satisfactory information is provided to participants to increase
perceptions of both self and response efficacy. The opposite is anticipated for gain oriented
message stimuli.
H7: Proactive coping influences adaptive protection motivations.
H8: Reactive coping influences adaptive protection motivations.
H9: The loss message condition moderates proactive coping, loss messages exhibit greater
influence on adaptive protection motivations than gain framed messages.
H10: The gain message condition moderates reactive coping, gain framed messages exhibit
greater influence on adaptive protection motivations than loss framed messages.

Protection Motivations
Boer and Seydel (1996) presented experimental subjects with information about mass
breast cancer screening (prevention) and tested a control group to assess protection motivation
theory models. The researchers found that experimental participants displayed a significantly
higher response efficacy, self-efficacy, coping appraisals and protection motivations to take part
in screening after exposure to the information than the control group (Boer & Seydel, 1996).

37

Arming consumers with useful information about how to cope with a recall event should produce
similar findings. Coping appraisals which offer ways to prevent undesirable health outcomes
have been demonstrated to influence protection motivations (Matos & Rossi, 2007). Coping
appraisals influence protection motivations and are based on whether perceived efficacy costs are
satisfied. The symmetrical relationships of gain and loss messages proposed in prospect theory
are proposed to produce opposite response effects on coping and behavioral intentions. The
participants call to action to react to the recall message content and avert illness is satisfied with
their coping appraisal. We believe that participants were more likely to take preventative actions
when the information presented to them is perceived as useful and acting on that information will
help them avert injury. If the information is perceived as less useful for helping the participant
avert injury, then they were less likely to adopt a coping strategy.
Four studies have produced favorable impacts on behavioral intentions by using factual
information to influence behavior in preventative contexts, (Milne, Orbell, & Sheeren, 2002;
Boer & Seydel, 1996; Seydel et al.; Steffen, 1990) however, no study has varied message content
informed by prospect theory to discover gain and loss message content's effects on protection
motivations. Milne, Orbell, & Sheeren (2002) determined that behavioral intentions increased
when subjects were exposed to motivational exercise intervention information that produced a
dual effect of increasing both threat and coping appraisals and behavioral intentions to engage in
exercise during a two-week three treatment experiment. Boer & Seydel (1996) conducted a
quasi-experiment using breast cancer leaflets and showed a significant impact for the leaflet
group on response efficacy, self-efficacy, and significantly higher behavioral intentions to
participate in mammography compared with a control group. These researchers did not find a
significant impact from the leaflet on perceived severity or perceived vulnerability to breast

38

cancer. Seydel et al. (1990) discovered similar effects among 624 subjects and determined that
threat appraisals were not significant predictors of behavioral intentions, but self-efficacy and
coping responses were significant predictors of behavioral intentions. Steffen (2006) used a
similar leaflet approach on testicular self-exams. Men who read the message treatments in the
leaflets demonstrated greater behavioral intentions to engage in testicular self-exam than those
who didn't read the leaflet. We expect the impact of each message type will demonstrate prospect
theory suppositions of risk aversion or seeking reflectivity, or opposite effects, around the
theoretical health reference point based on the gain or loss information supplied to subjects in
each trial condition.

Locus of Control
A recall message presents information to consumers about a potential health threat if a
damaged product is consumed. The consumer must assess this information to determine if the
threat is preventable through some action (s) on their part. In addition, the consumer may
interpret the recall threat information as outside their direct control and that reducing or
eliminating the threat is predicated on actions by others such as retailers or government agencies.
Protection motivation theory posits that self-efficacy is an important cause influencing
consumer's evaluation of threats because of specific abilities the consumer holds to reduce the
threats. Locus of control focuses about who is specifically responsible for reducing the risks
from a threat: the consumer him or herself or others (Bandura, 1977).
The consumer's perceived locus of control (e.g. self or others) suggests that their
behaviors to reduce those risks were either proactive or reactive in nature and are influenced by
whether they internalize or externalize the risks (Rotter, 1966). Internalized risks are directly

39

controllable by the consumer and are evaluated within the self-efficacy dimension by the
consumer's perceived willingness and ability to address those risks. Externalized risks are
viewed by consumers as being in the control of important others, such as retailers, and, it is not
the direct responsibility of the consumer to reduce the implied threats contained in the risk
(Harrington, 1996). Locus of control is theoretically useful for predicting why some consumers
faced with a risk from recalled products may take personal responsibility to prevent injury, while
other consumers think removing risks is the job of others. Locus of control is distinct from selfefficacy because self-efficacy includes an actionable ability to address the threat.

For this

reason, both constructs must be studied to better understand ways to improve consumer reactions
to recall messages. Workman et al. (2008) produced mixed results when measuring the impact of
locus of control on protective behaviors related to workers ignoring computer security measures
to protect their information. The researchers measured both objective measures (e.g. employee
compliance records) and subjective measures (e.g. employee self-reported survey results).
Subjective reports of locus of control had a significant impact on protection motivations, but
objective measures of actual performance of behaviors did not. This finding may be because of
subjects understanding what they should do does not match what they do when it comes to
protecting information. However, we feel that since food recalls present more serious personal
risks to consumer health our findings may be different. LaRose et al. (2008) measured subjects
based on personal responsibility for online safety risks. In four high or low involvement X high
or low self-efficacy groups, only the low involvement, low self-efficacy participants
demonstrated lower protection behaviors when told that protecting themselves was their own
responsibility. As expected, personal responsibility for online protection behaviors was greater
for individuals demonstrating high self-efficacy and high involvement. LaRose et al (2008)

40

found an interesting reversal for maintenance behaviors like updating virus protection definitions
in which people displaying high self-efficacy were more likely to react after a risk exposure than
performing routine maintenance. We predict a similar finding for consumer reactions to food
recall messages. Based on these findings, we speculate that loss oriented recall messages will
influence greater perceived internal locus of control and gain oriented messages will produce the
opposite effect by reducing the potential for maladaptive responses.
H11: Locus of control influences adaptive protection motivations in the loss condition.
H12: The loss message condition moderates locus of control, loss framed messages display
greater strength on adaptive protection motivations than gain framed messages.

Confidence in the Message Communicator
We will measure the impact of both message exposures (gain or loss) on consumer
perceptions of which message design is more useful for increasing protection motivations.
Aaker and Keller (1990) used a similar approach to assess two brand extensions in an
experiment. Each recall message framing type (gain or loss) may be perceived as a unique
communication product offered to customers. Newell et al. (2001) created a scale to measure
corporate credibility based on advertising messages. Their widely utilized scale devised from
prior research that measured consumer perceptions of businesses and/or the advertising messages
they presented to consumers (Goldberg & Hartwick, 1990; Mackenzie & Lutz, 1985; LaBarbera,
1982; Settle & Golden, 1974). The researchers found that their corporate credibility scale
provides a useful framework for measuring consumer perceived trust for a firm's communication
and expertise to produce those communications. As Newell et al. (2001) suggest in their study,
the analysis of the differences between consumer perceptions of trust and expertise may provide

41

better understanding for ways to deal with crisis situations like food product recalls. Lafferty
(2002) found that the credibility of different communicating sources based on Newell’s model
partially predicted the effectiveness of the advertising message based on the communicator
(corporation or sponsor). The role of the communicator in influencing specific behavioral
responses is not known but suspected to be similarly influential in explaining a portion of the
behavioral intention motivations. Sallam (2011) supported Newell et al. findings that corporate
credibility demonstrated a strong relationship to attitudes to the message, attitudes to the brand,
and behavioral intentions in a Saudi Arabian sample. In another study, Arora et al (2006) studied
message framing and source credibility’s influence on consumers who have largely ignored
appeals to exercise to improve health behavioral intentions. Recall responses by consumers
demonstrate a similar ambivalence to health appeals to modify behaviors. Arora et al. (2006)
found that highly credible sources demonstrate significant influence on behavioral intentions
compared to low credibility message framing. Subsequently, Arora (2007) tested the impact of
message framing, source credibility, and knowledge on attitude change and behavioral intentions
to buy teeth whitening products. The researcher found that positive framing was more effective
in changing attitudes and intentions than negative framing. Credibility of the communicator
demonstrated a significant relationship to product attributes. We similarly expect to see an
impact of message framing and confidence in the communicator on adaptive protection
motivations.

We plan to adapt the corporate reputation scale for our study to see the impact of

message framing (gain/loss) on the perceived credibility of the message communicator and
eventual consumer protection motivations.
H13: Confidence in the Communicator influences adaptive protection motivations.

42

CHAPTER 3
METHODS
Study Overview
The protection motivation theory hypotheses were tested with a three-condition
(gain/loss/control message) post-test only between subjects experimental design. Structural
equation modeling was used to measure the hypothesized dimensions. A pretest was conducted
of the post-test only experimental design to assess the veracity of the manipulations and
measurement model. After confirmatory factor analysis of the survey items, the main study was
conducted with a larger sample of participants using a between subjects post-test only online
experimental design.

Pretest Methods
Participants
We recruited participants for a pretest survey about shopping cart composition as the
basis for choosing a suitable recall message to include in the experimental design. The pretest
survey consisted of a sample of n=74 undergraduate students enrolled at a Midwestern
university. The pretest survey students received extra credit for taking part.
After analyzing the findings of the pretest survey, we recruited participants for a pretest
experiment which consisted of a convenience sample of size n= 99 of 200 undergraduate
students enrolled in a class at a Midwestern university. A response rate of about 50% of the
potential participants completed the online experiment. Pretest student participants received
extra credit for taking part in the pretest experiment.

43

Design
The pretest experiment used an online design to measure the hypothesized protection
motivation theory models by manipulating the effect of three conditions (gain/loss/control
message), using a classic prospect theory manipulation (Tversky and Kahneman, 1981).
Measures of susceptibility, severity, self and response efficacy, coping, locus of control, and
confidence in the message communicator were included. Participants were randomly assigned to
one of three experimental conditions for the pretest (i.e. control, gain, loss) message treatments.
Qualtrics software assigned the participants randomly into groups and ensured suitable cell size
by condition (control, gain, loss).

Measures
Participants in the pretest experiment were provided with a message stimulus about a
hypothetical current recall message (Figure 3b-d) in their randomly assigned condition
(control/gain/loss). The recall message was based on an actual recall message from Kroger.com
website involving Rocky Ford Farms cantaloupe in 2011. Respondents were asked to complete
48 survey items (Table 2) that measured protection motivations, susceptibility, severity, self and
response efficacy, coping, locus of control and perceived credibility of the message
communicator. After completing the survey in the gain or loss condition they completed 12
demographic measures (Table 3).
The pretest experiment allowed the research team to complete a confirmatory factor analysis to
assess the measurement model and adjust the experiment content prior to completing the main
study.

44

Data Collection Procedures
Participants received an email invitation to participate in the online experiment. Once
they clicked on the email link, they were assigned to an experimental group, and completed
informed consent and a screening question.

Participants were pre-screened prior to taking part

in the experiment to ensure they routinely make food purchases for individual consumption or
make food purchases for consumption by themselves and/or members of their household within
the prior six month period. If the participants did not routinely buy food products from grocery
stores, they were disqualified from the experiment and linked to a terminal thank you message in
the survey software. The disqualification rate during the pretest based on the routine food
purchases in the prior six months was 4% of respondents, n=4.
The participants reviewed the recall message in their randomly assigned group (i.e. gain,
loss, control). After reading their assigned recall message they were asked to complete the
survey. The survey measured the hypothesized dimensions of protection motivation theory and
prospect theory.

Findings
The pretest survey (Table 1) among 18-24 year old students at a Midwestern university showed
that 89% of 78 participants were familiar with Kroger grocery stores and therefore an
experimental stimulus using a retailer such as Kroger was likely to offer greater external validity
than creating a fictitious retailer for the stimulus. 65% of participants reported purchasing the
majority of groceries at mass merchant supercenters/hypermarkets such as Wal-Mart and Meijer,
while 29% reported buying most groceries at national grocers like Kroger.

45

Participants in the preliminary survey indicated that fruit was typically part of their
monthly shopping basket 78% of the time. Also, when presented with a shopping cart of
typically recalled products, 41% of the participants reported fruit as a ‘must have’ as their
primary choice, followed by deli meats at 24% of a typical basket. Therefore we used an
experimental recall notice involving fresh fruit to increase the external validity of the
convenience samples’ purchase behaviors. In addition, the participant preferences for buying
store brands versus national brands were roughly split at 55% and 45% respectively. Therefore
we used a fictitious owned Kroger fruit brand for the pretest and main study experiments recall
message stimuli in a further attempt to increase external validity (predominant owned brand
purchases) for study participants. The Kroger recall message adapted for the experiment was
the Rocky Ford Farms cantaloupe national recall caused by salmonella contamination that was
widely publicized in the national news media (Figure 3a). Kroger sold the impacted cantaloupe
in their stores and produced a recall notice on their website. ‘Rocky Ford Farms’ was changed to
‘Kroger Farms’ in the experimental stimuli to simulate a retailer owned brand for participants.
The remainder of the stimuli was identical to control message (Figure 3b).

Main Experiment Methods
Participants
After preliminary verification of the measurement model, survey items, and stimuli we
recruited participants for the main study. The main study tested a sample of participants based on
the 2010 census estimates of current adult (18+) population of the U.S. of 234,564 million
persons (Census.gov, 2011), with a required sample size with a confidence level of 95% and a
confidence interval of +/- 3.12 of 987 participants (Survey System, 2011). The targeted

46

population for the main study included typical food retail shoppers who might be impacted by
Class I product recalls in the United States. The demographics for sampling included both
genders, persons over 18 years of age or legal adults who are involved in food product purchases
from retail stores. To acquire an adequate sample size, we oversampled by approximately
21.7%, this plan was based on the recommendations from a ten-year study of online survey
response rates suggesting 30% oversampling (Sheehan, 2001). Recruitment of subjects for the
study was multi-faceted in an attempt to achieve the desired sample size. We recruited subjects
from the following sources for the experiment:
1. Using meta-tags in Qualtrics software so that potential participants doing searches for product
recall related information have the potential to find and participate in the study. The meta-tags
drew approximately 100 subjects to the survey website.
2. Amazon Mechanical Turk, a website offering a pool of paid experienced survey participants
was utilized for the majority of the sample. MTurk respondents were required to be U.S.
residents which had earned higher than average ratings in MTurk work assignments from
previous employers. MTurk subjects were compensated at the rate of $1 per completed and
verified survey. Workers reporting that they had not purchased food in a grocery store during
the prior six months were excluded from the study, n=48.
3. Contacting advocacy groups such as Stop Foodborne Illness (formerly Safe Tables Our
Priority), hereafter STOP, and soliciting participant lists to increase the number of participants
who have been diagnosed by a medical professional with food borne illness and those with
experience purchasing products that were later recalled.
Workers from MTurk chose to take part in the study based on reasons including: their
interest in the study subject matter (food recalls), compensation for completing the study, and

47

availability of the study in their potential work pool based on their experience/performance
rating on MTurk. Each respondent from STOP was sent a personalized experimental invitation
via e-mail to increase response rates. See samples of invitation message and reminder email in
Figure 2a-b. Reminder emails were sent to non-respondents at STOP after 14 days from the
initial questionnaire request with the goal of improving response rates.
The number of experiment invitations accepted was approximately 1260 with 987
surveys completed, or a response rate of 78%. Approximately 330 participants were randomly
assigned to each group (gain, loss, control).
Gender distributions on the Internet have demonstrated samples disproportionally
favoring male participants (Krantz & Dalal, 2000) and our study was an exception to this rule
with a good balance of male and female participants at 48% males and 52% females.
Experiments on the Internet may offer a wider age distribution than laboratory studies and more
closely approximate the targeted consumer population in the U.S. than convenience samples at
universities (Reips, 1996). Our survey had a reasonable distribution of ages (Table 3).
Similar to the pretest study, participants were pre-screened prior to participating in the
experiment to ensure they routinely made food purchases for individual consumption or made
food purchases for consumption by themselves and/or members of their household within the
prior six month period. The disqualification rate for the experiment was 4.9% or 48 participants.
The percentage was roughly one percent higher than the disqualification rate of the student
convenience sample utilized in the pretest.
Krantz & Dalal (2000) completed a meta-analysis of 20 online experiments and
discovered that excluding whites, no ethnic group exceeded 5% of Internet based experimental
subjects with excepting Senior et. al. (1999) study in which Chinese ethnicity was represented

48

between 5 and 10 percent of subjects. Our results were similar, however, Asian ethnicity
represented 34% of subjects, Caucasians comprised 52% and all other ethnicities were less than
5% of the sample pool.

Socioeconomic status was assessed utilizing two standard SES

indicators for health research including years of completed schooling, and household income
(Daly et. al, 2000). Sampling statistics were analyzed during the pretest and following main
study experiment to determine the mean, standard deviation, median, frequency and range to
determine if the desired sampling frame is achieved and is representative of typical adult
shoppers in the United States.

Design
The main study used an online experimental format to measure the hypothesized
protection motivation theory constructs by manipulating the message content for three conditions
(gain/loss/control message) on measures of susceptibility, severity, self and response efficacy,
coping, locus of control, and confidence in the message communicator. Participants were
randomly assigned to one of three experimental conditions for the pretest (i.e. control, gain, loss)
message treatments.
The experimental groups were divided into three groups: a control group, a loss group,
and gain message groups and subjects entered the experimental conditions by first reading a
stimulus for their respective condition (Figure 3b-d). Experiment participants each completed the
same survey items and demographic items regardless of their randomly assigned group. We
utilized a between subjects post-test only design to reduce the impact of exposing participants to
multiple message treatments (i.e. gain and loss).

Each group of randomly assigned participants

receiving the control, gain or loss oriented message were measured for the perceived impact of

49

the recall message on threat susceptibility, threat severity, self-efficacy, response efficacy,
coping appraisal, perceived credibility of the message communicator as independent variables
impacting the dependent variable protection motivation. Participants entered the experimental
conditions encountering a product recall message involving contaminated cantaloupe which
could potentially result in food borne illness if they failed to take action. The purpose of the
hypothetical recall message was to make the participant aware of the potential threat and
measure possible reactions to the threat in the present and similar threats in the future (behavioral
intentions or protection motivations).

Instrument
The relationship of the theorized dimensions to one another and subsequent protection
motivations were measured utilizing a structural equation modeling approach of prospect and
protection motivation theory (Tanner, Hunt, Eppright, 1991). The proposed model specification
measured the independent and dependent variables and the relationships between those variables.
A multi-group analysis approach was used in which approximately one-third of the participants
were randomly assigned to the gain oriented message stimulus, one-third of the participants were
randomly assigned to the loss oriented message stimulus, and one-third of the participants
comprised a control group.
The items used for the experiment are provided in Table 2, along with their source
instruments and reliabilities and the Likert scales utilized. A seven point Likert scale with a
neutral midpoint value (4) was used to measure all items. The experiment took participants ten to
fifteen minutes to complete after clicking the email hyperlink and completing informed consent.
Each measure utilized in the study was tested for internal consistency reliability using the

50

following measures: average inter-item correlation, average item total correlation, Cronbach’s
alpha coefficients and split half reliability. The validity of each instrument was examined for the
following including: content validity, internal validity, external validity, and discriminant
validity.
External validity was assessed to ensure the applicability of the findings to future tests of
recall message content. Statistical generalizability was assessed by drawing a random sample
from identified key independent variables (i.e. susceptibility, severity, coping) from the pretest
and online main study groups and comparing the means. Independent variables serve as a subset
of indicators for latent variables determining theoretical protection motivations with varying
weights or probabilistic influence on the risk averting behaviors (Lynch, 1982). Robustness was
assessed by comparing two types of consumers, students and non-students. If the means for
perceived credibility of the message communicator varied significantly between both groups,
then we cannot generalize across samples, thus presenting a threat to external validity (Lynch,
1982). These comparisons were made for the eight indicator variables that comprised perceived
credibility of the message communicator from the pretest group and main study groups. No
significant differences were found between the means of the pretest and main study groups.

Threat Assessment
After the participants read a stimulus similar to those presented in Figure 3b-d, the
severity of the potential threat from contracting food borne illness was measured by assessing the
perceived magnitude of harm, defined as a threat assessment, the participants expected if they
consumed the recalled product. The threat assessment items (Boer & Seydel, 1983; Rogers,
1983) addressed the proposed constructs of protection motivation theory: severity of the risk and

51

susceptibility to the risk. These constructs, while viewed as a composite variable, were separated
based on a stricter theoretical interpretation (Rogers, 1983). Severity (4 items) is defined as the
physical, psychological, or economic harm that may impact the consumer if the recalled product
is consumed. Each of survey dimensions for severity and threat assessment contribute to the
threat assessment. Susceptibility (4 items) measured and is defined as the likelihood that the
threat from the recalled product will impact the consumer or members of their household
personally. Susceptibility was measured by assessing the participant's perceived vulnerability to
food borne illness, perceived likelihood they will experience food borne illness if they eat the
recalled product, and the perceived relevance of the recall message for averting food borne
illness.

Perceived Efficacy
The next group of survey questions measured perceived efficacy (Sherer and Maddux, 1982).
Efficacy is defined as the perceived effectiveness of responding and participant’s physical ability
to respond, and perceived ease that a response will reduce the risk of injury from food borne
illness. Response efficacy (3 items) is the perceived degree that a response to the recall message
will reduce or eliminate the threats of food borne illness. We assessed whether participants
believe that the gain/loss oriented information will help them avoid injury and if the information
is adequate to help them avoid injury. Self-efficacy (5 items) measured the perceived ability to
respond to recall messages and whether they are capable of accessing recall information and
utilizing the information to avoid injury.

52

Locus of Control
The next group of survey items measured the participant's perceived locus of control (4 items)
for reacting to recalls based on the message content (Workman et al., 2008). Locus of control
items assessed whether participants perceived that recall message responses were within their
personal control as opposed to subordinate to the actions of others, such as corporations.

The

amount of coping response to the threat of food borne illness is contingent on perceptions of selfefficacy, response efficacy, and the locus of control. Locus of control items measured the
participant’s perceptions of whether it is within their direct control to reduce the injury or
whether important others are responsible for reducing the risk of injury.

Protection Motivations
We were interested in determining the anticipated longer term behavioral coping effects
of gain/loss framed messages on consumers. To assess these protection motivation behavioral
intentions we utilized 8 items adapted from Ho et al. (2005). We assessed the impact of
gain/loss/control messages on both maladaptive and adaptive coping responses. The between
subjects design measured the impacts of gain oriented or loss oriented messages on subjects to
determine the difference between message types and if one message type (e.g. control/gain/loss)
is more effective at influencing protection motivations. The control group will help us assess
whether message framing manipulations are useful for influencing protection motivation
behavioral intentions. The advantages of between subject designs are the anticipated reduction in
carryover, practice and learning effects common in within subjects experiments (Martin, 1996).

53

Perceived Credibility
The final group of 8 items measured participant's perceived effectiveness of the recall
message content for all treatment conditions (gain/loss/control) after the coping appraisal had
been measured. We will determine if perceived effectiveness of the message influences
consumer protection motivations for both the control, gain and loss groups as well. The
perceived credibility of the communicator scale is adapted from Newell and Goldsmith (2001)
(Table 2).

Data Collection
The study participants consisted of two groups including: an online panel sample
representative of typical U.S. consumers (MTurk) and a panel of members of an advocacy group,
Stop Food borne Illness (STOP). Experimental survey responses were recorded in Qualtrics
software application utilizing random assignment to the three test conditions: control, gain, and
loss message treatments. The responses from both groups were then tabulated using the
Qualtrics software and converted into a single SPSS data file. This SPSS file contained the
aggregated data set of all three experimental conditions and both groups delineated by
identifying variables (1=control, 2=gain, 3=loss and a separate variable to identify STOP
participants=1) which was then utilized to assess the distributions of the sample pools to survey
items.
Main study STOP experiment participants were sent an email invitation to participate in
the study, accompanied by an informed consent document. Participants clicked a hyperlink
directing participants to a Qualtrics based study website hosted at Michigan State University
Department of Marketing. The sample email invitations and informed consent email are

54

provided in the Figure 2a-b. Main study MTurk participants responded to an eligible work
assignment which contained the same text as the email invitation and informed consent
confirmation. After participants clicked on the hyper-link they were directed to an experimental
landing page on the Qualtrics study site. This page contained pre-screening questions including
an assurance that the participant routinely purchased food items from grocery stores in the past
six months and confirmation that the subject thoroughly read the message stimulus prior to
advancing to the survey questions. Participants in the experiment were provided with a message
stimulus (Figure 3b-d) in their randomly assigned condition (control/gain/loss) about a
hypothetical current recall message of private branded cantaloupe at a national grocery store and
then be asked to complete 48 survey items (Table 2) that measured protection motivations, locus
of control and perceived credibility of the message communicator. After completing their survey
in the gain or loss condition they then completed 12 demographic measures (Table 3).
you message screen ended the survey.

55

A thank

CHAPTER 4
ANALYSIS

Statistical Analyses
Because of the challenges of low consumer responses to recall messages, it is
important that researchers develop and test reliable measures for examining the hypothesized
variables. The primary purpose of this study is to understand the role varied recall message
treatments have on consumer protection motivations. Different recall message content may
influence consumer tendencies to protect themselves from potentially dangerous food borne
illness events based on reactions to cues presented in recall messages. An important
contribution of the study is to understand the psychometric properties of theorized variables and
how they influence the factor structure of the measurement scales used in the study.
The frequency characteristics, measures of central tendency, and variance
statistics were analyzed by descriptive statistics techniques in SPSS software. Four items in the
survey used the technique of reverse coding and were recoded before analysis. The results of
this study are discussed in the following sections: 1. demographics of the sample, 2. descriptive
statistics of the survey items, 3. measurement model characteristics and findings, and 4.
structural model analysis.

Demographic Data Analysis of the Sample
The first phase of the experiment involved a sample from Amazon Mechanical Turk (MTurk)
and was collected from May 3 – May 17, 2012 using Qualtrics survey software. Participants
were randomly assigned to one of three experimental conditions (control, gain, loss). 1038

56

surveys were started and 818 were completed, indicating a response rate of 79%. The
experiment and survey took on average 10 minutes for the MTurk participants to complete. The
second phase used a sample from an online food borne illness advocacy group, STOP Food
borne Illness (STOP) and was collected from May 4 – May 31, 2012 using Qualtrics and random
assignment to the three experimental conditions. 222 surveys were started and 121 were
completed for a response rate of 55%. The experiment and survey took on average 14 minutes
for the STOP participants to complete. We suspect the fact that STOP participants were eligible
for a random drawing as opposed to guaranteed compensation to complete the survey
substantially lessened the response rates for STOP participants.
The demographic characteristics of each sample are provided in the Table 3. The two
samples (MTurk, STOP) were combined for the measurement model analysis, but several key
points about differences between sample pools are noteworthy.
The MTurk sample was closer to U.S. census gender distributions of 50.8% females
(census.gov, 2012) at 52% male and 48% female, however the STOP sample consisted of a
disproportionately larger sample of females at 75% female and 25% male respondents. The
MTurk sample consisted of mainly younger individuals with 52.8% or respondents under the age
of 34 compared to the STOP sample at 16.8% under 34. The STOP sample was made up of
mostly Caucasians ( 91%). The MTurk sample had a disproportionate number of Asian
participants, 38.8% compared to U.S. distributions for Asians in the United States at 4.8 percent
(census.gov, 2012). Recall experience, defined as buying an item that was later recalled was
higher for the STOP sample at 51% versus 40% for the MTurk sample. Diagnosis of food borne
illness was 6% higher for the STOP sample than the MTurk sample. Annual income was lower

57

for the MTurk sample with 64.6% of respondents earning less than $40,000 a year compared
with the STOP sample in which 23% of respondents earned less than $40,000 a year.

Descriptive Statistics of the Survey Items
Table 4 shows the distributions for the survey items used for this study. A seven point
Likert scale with a neutral midpoint value (4) was used to measure participant responses.
Several of the item responses are interesting for future research to assess message manipulation
in PMT. One item response pattern (Table 4, #4) showed that a greater percentage of
respondents disagreed than agreed the recall message stressed the health benefits of not eating
the cantaloupe (46.6% disagreeing and 39% agreeing, M=3.77, SD=1.846). Health benefits in
the recall message may influence perceived effectiveness of the recall message and should be
explored in future research. By contrast, respondents seemingly understood the message about
the risks of eating the cantaloupe (Table 4, #5) with (88% agreeing and 8% disagreeing, M=5.8,
SD=1.319).
Responses to the locus of control items displayed an interesting phenomenon often
encountered in individualistic cultures, in which, responsibility is viewed within the individual’s
control to prevent illness (Rotter, 1966). 83% of respondents (Table 4, #14) reported that
preventing illness from recalls was their responsibility as opposed to agreeing that it was others
(e.g. retailers) to prevent illness from recalls at 39%. 47% of respondents (Table 4, #15)
disagreed that it was the responsibility of others to prevent injury from recalls. Participants feel
that they are personally accountable to take action to ensure the safety of their food and react to
recall messages they read.

58

Participants reported that factors which might influence maladaptive responses such as
invulnerability to illness because of their genetic heritage, “luck”, and no prior experience with
food borne illness were not correct assumptions for avoiding potential illness. 76.3% disagreed
with power of their genes (Table 4, #28) to protect them, 53% refuted their luck (Table 4, #29),
and 65% discounted their limited prior experience (Table 4, #30), with food borne illness. This
finding challenges the assumptions in prospect theory of positivity biases in subjects. The
finding also highlights peoples’ understanding of the potential for foodborne illness when
defective food products are eaten.
On the other hand, after reading recall messages respondents strongly agreed with
statements measuring protection motivations. In fact, 92.5% of respondents (Table 4, #34)
reported that they will use recall message information to lessen the risk of illness from eating
recalled products. This item demonstrated the highest agreement in the sample pools. 91.1% of
respondents also reported agreement that information in recall messages provides a means to
reduce the risk of serious illness (Table 4, #35).
The MTurk survey invitation provided participants with an opportunity to provide
researchers with feedback about the study. Supplementing strong levels of agreement were
many favorable comments from respondents about the importance of raising awareness of recall
messages in the United States. Participants view recall messages as an important way to reduce
illness and as a tool for educating consumers. 143 MTurk respondents completed comments
about the study. 130 of these responses were favorable and 13 were negative. Some examples
of positive comments and all negative comments are provided in table 14.

59

Measurement Model Characteristics
Structural equation modeling allows researchers to measure the relationships between
measurement components and structural relationships of components to one another. The
method models independent and dependent constructs, including latent or unobserved variables.
Structural equation modeling first identifies a measurement model through the combination of
empirical measurement indicators into unobserved theory driven latent variables. Confirmatory
factor analysis is a statistical tool to test the measurement model assumptions to the empirical
data (Bollen & Lennox, 1991). Measurement indicators may be reflective or formative. For
reflective indicators, causality flows from the latent variables, and error is measured at the item
level. Reflective indicators are expected to be correlated. Formative indicators differ from
reflective indicators because causality flows from the indicators to the latent construct and error
is measured at the latent construct level. Formative indicators are not expected to be correlated
(Bollen & Lennox, 1991). This study utilized a formative modeling approach for the following
reasons:
1. Indicators are defining characteristics of the construct
2. Changes in the indicator should cause changes in the construct
3. Changes in the construct should not cause changes in the indicator
4. Indicators need not be interchangeable
5. Nomological net for the indicators may differ
6. Indicators do not need to share a common theme (Jarvis et al., 2003, p. 203)
Structural equation modeling allow researchers to examine the direct and indirect effects
of exogenous (measurement) variables and endogenous (latent) variables. In structural equation

60

modeling the exogenous variables are similar to independent variables which influence the
relationships between the endogenous variables.

Pretest Analysis
A pretest experiment was conducted as a preliminary test of the measurement model
before the main study.

To test the pretest measurement model, the three conditions were added

into a single sample to ensure measurement equivalence. 38 items were included in the pretest
study and were the basis for nine latent variables theorized to exist in the model. These included:
susceptibility, severity, response efficacy, self-efficacy, coping, maladaptive protection
motivation, adaptive protection motivation, locus of control, and confidence in the message
communicator. The latent factors are assumed to account for some part of the observed variation
with the residual error term which is unmeasured. The 38 items or observed variables in the
measurement model assume covariance’s may be explained by the underlying latent factors.
This theoretical recall protection motivation model is constructed from many measurement scale
sources and is untested in the literature. As a first phase of the study we used exploratory factor
analysis (EFA) with the purpose of discovering a suitably parsimonious model. After ensuring
as few survey items as possible, we then conducted a confirmatory factor analysis. One goal of
exploratory factor analysis is to reduce participant fatigue by reducing the number of survey
items. Another goal is to remove less effective survey items from the design.

Table 6 contains

2

the findings of the CFA, X = 839, degrees of freedom 239, p<.045 The critical ratio or Wald test
determined by dividing the estimate by the standard error more than 1.96 or higher indicated that
all factor loadings were significant.

61

Before performing the main study, a survey item was added to the self-efficacy construct
to increase scale reliability and compensate for the poor loading of one self-efficacy item (.41),
selfeff2 which showed an effective loading of .74 for the self-efficacy construct (#20, Table 7).
Due to homogeneity of participants, small sample size and cell depth of the pretest sample, all
items were retained for the main study experiment. Retaining poor loading (<.50) items based on
the pretest sample size n=99 was prudent as, with a single exception, formative indicators
improved in the measurement model of the main study. Poor loadings in the pretest
measurement model and main study loading are as follows: susceptibility 1=.52 (pretest) .64
(main study), locus of control 2=.58 (pretest) .69 (main study) locus of control (SELFaccountability) .34 (pretest) .44 (main study), maladaptive protection motivation 2 =.53 (pretest)
.73 (main study), self-efficacy 1=.56 (pretest) .75 (main study), adaptive protection motivation
2= .49 (pretest) .62 (main study), and coping 2 .58 (pretest) which decreased to .57 in the main
study. Locus of control OTHERS responsibility were removed from the study based on failure
to load on the locus of control construct.

Main Study Analysis
The main study measurement model assessment was conducted in a manner similar to the
pretest model assessment. An EFA of all measurement items was conducted before a CFA scale
reduction to improve overall model fit within the larger collected sample (N=921. Χ² = 4607,
degrees of freedom = 666, p<.001). All unstandardized estimates in the aggregated samples
(MTurk and STOP) measurement model were significant, p<.001. To assess the structural
component of the model, items with poor loadings were removed. The revised CFA
measurement model may be found in Figure 5.

Model fit improved dramatically between the
62

main study EFA analysis, (χ² = 4607, degrees of freedom = 666, p<.001) and the reduced model
2
main study CFA, (X = 1081, degrees of freedom = 305, p<.029) (Table 6). Similar

improvements were noted in the more parsimonious CFA measurement model, SRMR = .0484,
RMSEA = .053, GFI =.922, AGFI = .897, CFI = .948 compared to, SRMR =.080, RMSEA
=.078, GFI = .77, AGFI =.73, CFI = .812 for the full EFA model.
The results of the main study EFA yielded 14 items with poor loadings. Thus, these 14
items were removed, and 25 were retained to form the reflective indicators of the reduced CFA
measurement model. Items deleted include (table 7): Susceptibility 1, Severity 1, Self-efficacy
3, Coping 1-3. The Coping items formed a sub-dimension of coping which is best described as
proactive coping compared to reactive coping (the retained items included reactive coping 46).Other items removed include: Adaptive protection motivation, Locus of control, and
Confidence in the communicator 1-4.
The measurement model was assessed for internal consistency using Cronbach’s alpha.
The alphas for the CFA are presented in Table 10. Convergent validity is demonstrated when
each of the measurement items loads with a significant value on its latent construct. Convergent
validity was measured by using the standard suggested by Fornell & Larcker (1981), loadings
over .50 and the AVE explained was greater than the average variance unexplained or indicated
within measurement error. Discriminant validity, or the measurement of how constructs differ
from one another without sharing variance between several constructs, was calculated by
comparing the square root of the AVE to the item to construct correlations. Discriminant
validity is established when the measurement items show a suitable pattern of loadings based on
theoretical assumptions of the assigned factors (Gefen, 2005). Factor loadings which are less
than .7 suggest that greater than 50% of the variance in an observed variable is explained by
63

factors different from the construct to which the indicators are theoretically related (Farrell &
Rudd, 2009, p. 3) The underlying items support the theoretical constructs with the exception of
the severity 3 variable which had a loading of .63, but was kept for analysis because of its
theoretical relevance.

Confirmatory Factor Analysis
The factor structure of the 27 item post-EFA scale was examined. All but three of the items
correlated >.30, indicating factorability. The remaining items were retained for theory
assumptions. Next, the Kaiser-Meyer-Olkin measure of sampling adequacy was .93, well above
2
the recommended value of .6. Bartlett’s test of sphericity was significant X (378) =15087,

p<.001. The communalities were all above .5, indicating that each item shared common variance
with other items. The anti-image correlation matrix diagonals were all above .5. Given these
findings, confirmatory factor analysis was a viable option for the sample.
Principal component analysis was used to examine the ten factor model: susceptibility,
severity, self-efficacy, response efficacy, proactive coping, reactive coping, locus of control,
confidence in the communicator, adaptive and maladaptive protection motivations. The factor
structure was examine with direct oblimin rotation, D=0. Initial Eigen values for the ten factor
solution based on the theoretical model explained 77.27% of variance. Eight of the ten Eigen
values explained 69% of the variance with values greater than 1.0. Each of the final two factors
in the model explained approximately 4% of the variance and were retained for theoretical
purposes.

64

Manipulation Check
As suggested in Perdue & Summers (1986), study we have two primary goals in
performing a manipulation check of the gain and loss oriented message content:
1. Ensure that treatment manipulations are related to the measures of the latent variables they are
designed to influence.
2. Assess whether changes among related but independent (control, gain, loss) conditions are
statistically significant.
We carried out these goals by comparing the means of the manipulation check items and
completing an ANOVA on these for the aggregated sample. The dependent variables are
normally distributed and verified with q-q plots. We determined whether each message group
(e.g. gain, loss or control) differed significantly on protection motivations. The following
procedures were used to perform the manipulation check:.
1. The manipulation check pretest consisted of four items adapted from Meyers (2010), (Table
2). These items measured risks of consuming the cantaloupe. To ensure participants perceived
the expected differences between each message condition, responses were compared between the
control, gain, and loss groups from the combined MTurk and STOP samples.
2. To discover the strength of the manipulations, a one way analysis of variance was performed
to measure perceived differences among the three conditions. The findings of the manipulation
check are found in table 5. Two of the manipulations check items showed there were significant
differences between the three experimental groups: “the recall message emphasized what might
happen in the future if I ate the cantaloupe”, F=6.584, p<.001 and “the recall message
emphasized the health benefits of not eating the cantaloupe”, F=7.558, p<.001.

65

Next, we examined our manipulations with a moderation test in AMOS on the
hypothesized model to determine the effect of the predictor variables upon other predictor
variables. The predictor variables are continuous and the moderator (gain, control, loss)
conditions are categorical. The full model was tested and each insignificant path’s regression
weight shared among all three continuous moderators (control, gain, loss) were deleted
successively until only significant predictors in any of the three experimental conditions (control,
gain, loss) remained for moderation analysis.

We regressed each of the outcome variables on

the predictor variables for each of the experimental conditions. We then compared the
unstandardized regressions for each of the three regressions. Based on the differences in the beta
weights, there is evidence for moderation (Table 5B). The unconstrained model comparison to
2
the fully constrained model chi-square difference test, X (16)=27, p<.041, indicates the

experimental conditions are not invariant, allowing us to reject the null hypothesis that there are
no differences between the three conditions (control, gain, loss). In addition three latent
variables provided evidence for moderation effects among the three experimental conditions
2
(control, gain, loss), including proactive coping, X (17)=54, B=.138, p<.001; locus of control,
2
2
X (17)=55.1, B=.067, p<.001; and reactive coping X (17)=53.9, B=.259, p<.001.

Common Method Bias
Common method variance in structural equation modeling occurs when correlations are not
related to actual relationships between the variables, but since they were measured by the same
method. In this study, each participant completed the same survey after reading the recall
message. To analyze common method bias, we used the Multi-Trait- Multi-Method approach.
First, we analyzed the data utilizing Harmon’s single factor test. We completed a factor analysis
66

of the EFA main study model in SPSS combining all measurement indicators into a single factor
solution with no rotation. The variance in the measurement model described by a single factor
produced an Eigenvalue of 39 with 99.99% of total variance explained. The CFA model was
then tested using Harmon’s single factor test by combining all measurement indicators into a
single factor with no rotation. The amount of variance in the parsimonious measurement model
described by a single factor produced an Eigenvalue of 10.56 and 37.71% of total variance
explained. Next, we performed a common latent factor test in AMOS and created fixed value
regression lines to each of the CFA indicator variables in the parsimonious measurement model.
The common regression weight was B=.37, which was then squared to determine the common
variance of the measurement model , or .137 shared variance. Finally, we performed a marker
variable test by adding a latent construct and indicators from the dataset, but not part of the
model and regressed the common factor construct on the revised model. The market variable
2
test, B=.25, B =.0625, reducing the common shared method variance of the model to

approximately 6%, and lower than the common criteria of <10 %. We can conclude that 6%
common method variance created from using the same protection motivation indicators in each
of the three message frames to measuring constructs influenced minimal bias in our results.

Formative Latent Factor Structural Test
Further manipulation check testing used structural equation modeling and included a
multiple regression manipulation check of the independent variables influence on the construct
of the four manipulation check indicators. The findings are presented in figures 6a-c. While all
variables were retained for analysis, the control manipulation produced a greater impact on
2
subjects compared with the loss or gain oriented message conditions as evidenced by the R =.30,

67

2
compared to R =.23 for the gain and loss model. This lower explanatory power of both

experimental conditions compared to the control provides further support that the models are not
invariant and message manipulations influence perceptions of PMT variables. Subjects’
manipulation perceptions did not vary between gain and loss messages. In the gain condition,
items for response efficacy significantly influenced the manipulation construct. In the loss
condition, severity, self-efficacy and proactive coping were significant. In the control condition,
susceptibility, self-efficacy, confidence in the message communicator were significant. This
finding suggests that gain and loss conditions may alter consumers’ perceptions of the recall
message in different ways. Self-efficacy is significant only in the control and loss condition,
suggesting a greater influence on the consumer’s willingness and ability to respond in the
presence of the fear appeal, or lack of information in the control condition. This finding may also
be supported by the significance (p<.05) for the severity variable in the loss condition, which
may also be related to the fear appeal.
For the manipulation check in the gain condition only response efficacy was significant.
This finding is logical because the message focuses on those who will not become ill and thus is
related to taking action. Recall messages that stress how many people will not become ill may
increase consumers’ perceptions of response efficacy, or the effectiveness of responding. By
contrast, recall messages highlighting how many people will become ill may increase consumer
perceptions of severity and self-efficacy. Messages without gain and loss manipulations
influence susceptibility, self-efficacy and confidence in the communicator. The suppression of
confidence in the communicator in the gain and loss conditions indicates including probable
percentages of outcomes does not instill confidence in the communicator. However, these
probabilistic cues may influence protective behavior perceptions and taking action to reduce the
68

risk of illness. The goal of examining each of the three groups in a reflective model using SEM
shows whether people perceive the constructs differently in each condition. The variation in
significance is a good predictor that the gain, loss, and control conditions are interpreted
differently by people.

Hypothesis Testing
We utilized a multivariate analysis of variance (MANOVA) prior to testing the path
model. MANOVA allowed us discover whether the hypothesized independent variables have an
effect on the two dependent variables: adaptive and maladaptive protection motivations in each
experimental condition (control, gain, loss).
MANOVA is useful for measuring the combination of the independent variables main
effects on the dependent variables and provides more meaningful results than considering each
dependent variable separately as in an ANOVA test. MANOVA also examines the intercorrelations of the dependent variables in the model. To ensure the homogeneity of the
covariance matrices across groups we conducted a Box’s M test.. Box’s M test was not
significant, confirming the assumptions of MANOVA, F.623, df(1)=6, df(2) 208413, p<.712.
A one-way MANOVA showed a non-significant multivariate main effect for
susceptibility on adaptive and maladaptive protection motivations. A one-way MANOVA
demonstrated a significant multivariate main effect for severity on adaptive and maladaptive
protection motivations, Wilk’s λ = .988, F(2, 910) = 5.35, p<.01, lending support for hypothesis
two. A one way MANOVA demonstrated a significant multivariate main effect for self-efficacy
on adaptive and maladaptive protection motivations., Wilk’s λ = .641, F(2, 910) = 254.64,
p<.001, lending support for hypotheses three and four. A one way MANOVA demonstrated a

69

non-significant multivariate main effect for response efficacy on adaptive and maladaptive
protection motivations. A one-way MANOVA showed a significant multivariate main effect for
reactive coping on maladaptive and adaptive protection motivations, Wilk’s λ = .925, F(2, 909) =
15.25, p<.001, lending support for hypothesis eight. A one-way MANOVA showed a significant
multivariate main effect for proactive coping on maladaptive and adaptive protection
motivations, Wilk’s λ = .968, F(2, 909) = 36.985, p<.001, lending support for hypothesis seven.
A one-way MANOVA demonstrated a significant multivariate main effect for locus of control
on protection motivations, Wilk’s λ = .988, F(2, 910) = 5.365, p<.01, lending support for
hypothesis eleven. A one-way MANOVA demonstrated a significant multivariate main effect
for confidence in the message communicator on protection motivations, Wilk’s λ = .901, F(2,
910) = 50.10, p<.001, lending support for hypothesis thirteen.
Levene’s test for equality of error variance was non-significant for adaptive and
maladaptive protection motivations as expected, allowing us to reject the null hypothesis,
confirming the findings of the moderation test.

Path Analysis
Next, we examined the structural model using path analysis and structural equation
modeling in AMOS to analyze the multiple regression model. From this analysis, we hope to
better understand the relationships within the formative model from each latent predictor
construct influence on maladaptive and adaptive protection motivations independent of shared
variance between them. All PMT hypothesized independent variables were allowed to co-vary
freely for this analysis and a three group model (control, gain, loss) were tested. The fit statistics
2
for the path model X (3)= 9.243, p<.026 indicate acceptable fit: GFI = .99, AGFI=.89,

70

SRMR=.031, RMSEA=.048. However, because of the multi-group experimental design and the
significance of path coefficients, the analysis will focus on individual paths and their requisite
beta weights. The critical ratios in excess of -1.96 at p<.05 and 1.65, α = .10, for the predictors
effects on adaptive and maladaptive protection motivation were examined in the analysis. We
controlled our analysis for each experimental condition (control, gain, loss) as categorical
grouping variables.
The path model tests the relevant strength of fit of the hypothesized model to the latent
constructs. Path coefficient values represent the strength of the correlations between the
independent and dependent variables. The path model tested in this study is a multiple
regression path analysis through structural equation modeling in which we assume the
independent variables are correlated.

Hypothesis Results
The table summarizing all hypotheses tests is presented in Table 13 and path coefficients
are presented in Figures 7a-c. Hypothesis one predicted that perceived susceptibility would
influence maladaptive protection motivations in the control condition. We theorize that
susceptibility effects are suppressed when message framing is present. The relationship between
perceived susceptibility and maladaptive protection motivations was not significant. Hypothesis
one is not supported.
Hypothesis two predicted that severity would influence maladaptive protection
motivations in the control condition. Similar to hypothesis one, we theorize that message
framing suppresses the effects of perceived severity on maladaptive protection motivations.
Hypothesis two is supported, B=-.226(.091), p<.05.

71

Hypothesis three predicted that self-efficacy, or the willingness and capability to respond
would influence maladaptive protection motivations. In all three experimental framing
conditions the relationship between self-efficacy and maladaptive protection motivations was
significant. Control B=-.579 (.049), p<.001; Gain B=-.621(.048), p<.001; Loss B=-.688(.045),
p<.001.
Hypothesis four predicted self-efficacy would influence adaptive protection motivations
in the gain condition. We anticipated the impact of gain message framing influence on
increasing participant perceptions of willingness and capability to take action to reduce risk.
Hypothesis four is supported, B=.091 (.027), p<.001.
Hypothesis five predicted that response efficacy would influence maladaptive protection
motivations, similar to self-efficacy, in all message treatments. No significant relationship
between response efficacy and maladaptive protection motivations existed. This is probably due
to the fact that the hypothetical recall message reduced the perception of favorable outcomes
from responses on which the theoretical variable contends. Hypothesis five is not supported.
Hypothesis six predicted that response efficacy would influence adaptive protection
motivations in all message treatments. No significant relationship between response efficacy and
adaptive protection motivations occurred. Hypothesis six is not supported. Similar to hypothesis
five, there may be a dampening effect on perceptions of favorable outcomes from responses to
recall messages when they are hypothetical situations and not actual risk events. To better
understand the perceived benefits of recall messages responses, it may be necessary to test actual
recall event responses in real time.
Hypothesis seven predicted that proactive coping would influence adaptive protection
motivations across all three message treatments. Hypothesis seven was supported, Control

72

B=.127(.047), p<.01; Gain B=.095(.043), p<.01; Loss B=.195(.042), p<.01. Making a plan to
avert illness in the future demonstrated the highest regression weight in the loss condition. This
suggests that loss framing may be useful for influencing ongoing behavior change in recall
message responses. This logic supports the participant’s maintenance of their health reference
point.
Hypothesis eight predicted that reactive coping influences adaptive protection
motivations across all message framing treatments. Hypothesis eight is supported, Control
B=.183(.056), p<.001; Gain B = .357(.053), p<.001; Loss B = .223(.052), p<001. The gain
message frame revealed a demonstrably higher regression weight. This finding may indicate that
when participants weigh the risk that most people will not become ill, they are more likely to
respond to the message when they read it to achieve similar results. This logic supports
participant’s maintenance of their health reference point.
Hypothesis nine predicted that the loss message frame moderates pro-active coping on
2
adaptive protection motivations. Hypothesis nine is supported. The X difference test showed

evidence that the loss condition moderated proactive coping on adaptive protection motivation,
with a chi-square difference threshold, 54.2 (17), p<.01. For people predisposed to proactive
planning to avoid risks, loss framing appears to be more influential for motivating behavior
change.
Hypothesis ten predicted that the gain message frame moderates reactive coping on
2
adaptive protection motivations. Hypothesis ten is supported. The X difference test

demonstrated evidence of the gain condition moderating reactive coping on adaptive protection
motivation, with a chi-square difference threshold, 53.9(17), p<.05. For participants faced with

73

gain message framing, there appears to be a greater desire to react the message immediately to
receive the implied gains of avoiding illness.
Hypothesis eleven predicted that locus of control would influence adaptive protection
motivations in the loss frame. Hypothesis eleven was supported, B=.144 (.048), p<.01. For
participants faced with probable losses, the influence of internal locus of control appears relevant
to taking action to avoid the risks.
Hypothesis twelve predicted that the loss message frame moderates locus of control on
2
adaptive protection motivations. Hypothesis twelve is supported. The X difference test

demonstrated evidence of the loss condition moderating locus of control on adaptive protection
motivations, chi square difference threshold, 55.1(17), p<.01. When events with probable losses
are presented to participants, the relationship of internal locus of control to taking action to avoid
risks appears to be an important consideration.
Hypothesis thirteen predicted that confidence in the message communicator would
influence adaptive protection motivations in each message treatment. Hypothesis thirteen is
supported, Control B =.277 (.045), p<.001; Gain B =.195(.045), p<.001; Loss B = .250(.043). It
is interesting to note that the regression weight was slightly higher in the control condition
compared to the loss condition. This finding is interesting and suggests that probabilistic
message manipulations framed as gains/losses may reduce subject confidence in the message
communicator. This is likely due to unfamiliar recall messages explored in this experimental
design, and additional cognitive processing they elicit in subjects. Attribution theory may also
explain this phenomenon, in which subjects attribute blame to the source communicator (Shaver,
1985). To better understand this phenomenon however, repeated measures or a longitudinal
study would be required to measure learning effects from novel message manipulations.
74

To assess the models predictive power of maladaptive and adaptive protection
2
2
motivations, the R was calculated for each condition (Table 14). R Measures the variation in

the dependent variables explained by the model. The estimates for the dependent variables
2
2
squared multiple correlations or R = .539 for adaptive protection motivation and R =.415 for

maladaptive motivation in the control condition. In the gain condition the hypothesized
relationships of the protection motivation variables to adaptive and maladaptive protection
motivations were R

2=

.557 and R

dimensions accounted for R

2=

2=

.471 respectively. In the loss condition, the hypothesized

.605 and R

2=

.511 of the variance, signaling the loss condition

displayed the greatest potential influence of the independent variables to explain protection
motivations.
This parallelism in adaptive protection motivation suggests that specific paths show
different influences to achieve a similar result through the predictor variables. For example,
coping strategies, internal locus of control and confidence in the message communicator are
instrumental in explaining the variance in adaptive protection motivations in the loss condition.
In the gain condition, however, the internal locus of control variable is suppressed, while selfefficacy influences adaptive protection motivations in addition to coping strategies and
confidence in the communicator. This finding is interesting and suggests that different cognitive
processes are involved in formulating behavioral intentions when gain/loss message frames are
present in recall messages. In the gain condition, both locus of control and self-efficacy
influence maladaptive protections motivations, likely by eliciting fear and increasing the
intentions to take action. The moderating effects of coping strategies and locus of control

75

perceptions when people face risks needs further study. Likewise, the tendency of people’s
willingness and capability to react to recall risks when presented with gain message content also
needs to be better understood.

76

CHAPTER 5
DISCUSSION AND CONCLUSION

Discussion
The goal of this study was to examine framing effects in food recall communications to
discover their impact on protection motivation theory models. Specifically, we wanted to
measure the effects of theorized constructs on adaptive and maladaptive protection motivations.
To carry out this goal, we used a between subjects experiment that tested subjects by having
them read a recall message from a retailer website and then answering a survey. Three
conditions were tested: a control, a gain framed message, and a loss framed message. The gain
and loss message frames used a prospect theory manipulation in which the impact for risk of
illness was identical, but the wording was different. A follow-up survey was utilized to measure
differences between subjects’ perceptions of the threat. We measured the effects of efficacy,
coping, locus of control, perceived credibility on protection motivations.
Protection motivation theory predicts subject behavior changes when faced with risk(s)
(Rogers, 1983). Prospect theory focuses on consumer decision making which concentrates on
increasing gains. Integrating protection motivation theory and prospect theory to study
responses to recall messages was a novel approach used in this research to predict how consumer
reactions to recall messages differed when framed with gain and loss message manipulations.
Tversky and Kahneman (1981) demonstrated that when participants are faced with the
same problem, presented in different ways, that their preferences change. Our study measured
subjects faced with the possibility that 80% of people will not become seriously ill versus 20% of
people will become seriously ill. Gain oriented messages focus on the advantages of greater
77

percentages of people not becoming ill and loss oriented messages stress the disadvantages of
people becoming ill as a percentage.
Previous findings from prospect theory suggest that framing is a powerful motivator of
consumer perceptions. However, as O’Keefe & Jensen (2006) suggest, challenges exist for
influencing behavior change in subjects with messages containing prevention and detection
tactics. Most behavior changes are short term unless losses occur periodically to change the
consumer’s equilibrium or reference state. Protection motivation theory tries to understand
subject decisions that lead to protective behaviors. Prospect theory examines subject choices
when facing risks. The bridging of protection motivation and prospect theories to address food
recalls, a serious risk, is a logical method for examining decision making under risk. Our
research desires not only to better understand the impact of message frames on increasing
adaptive protection motivations, but also the influence of the message frames (control, gain, loss)
on maladaptive protection motivations.
Recall messages fulfill a dual role of detection and prevention. A business issuing a
voluntary recall message wants the consumer to search their home (detection) to find out if they
own the recalled product. Next, a consumer will engage in prevention behaviors by properly
handling the recalled product. This includes cooking, returning, or disposing of the product.
Detection and prevention are not mutually exclusive, but interdependent with detection leading
to prevention. Detection is necessary but not sufficient for preventive behavior to occur.

We

suspected that when subjects weighed negative framing for avoiding risk of illness they are
influenced more than gain framed participants. This effect produced by the impact of
probabilistic statements of sure losses and likely gains controlled in the message content.
Preventing injury highlighted in sure loss statements influences participants behavior more than

78

the preventive statements of likely gains. This result is likely due to the immediate nature of the
threat by stating a sure loss. Our research study supported this finding in several ways that
summarized below. We will first review the results of message framing on maladaptive
protection motivations followed by adaptive protection motivations.

Maladaptive Protection Motivations—Loss Condition
We found that, when self-efficacy increases, maladaptive protection motivations
decrease. Participants’ reading a message implying that some people become ill (risk of loss)
influences a willingness to react to the message and avoid injury. The loss framed message is
effective at decreasing the potential to not take action. This finding supports prior PMT research
where loss frames influenced a reduction of maladaptive protection motivations through selfefficacy and response-efficacy constructs (Rippetoe & Rogers, 1987; Boer & Seydel, 1996;
Milne et al., 2006). No relationship existed between response efficacy and maladaptive
protection motivations in our study. Self-efficacy produced a significant influence on
maladaptive protection motivations. This suggests that participants sensed a capacity to respond
to a threat but are uncertain if their response will be effective in mitigating the threat of food
borne illness. This finding is surprising and may indicate that loss condition messages created a
stronger influence on self-efficacy, but the participants felt that responses would not protect them
from illness. Prior consumption of contaminated food, doubt of exposure, or other causes may
explain this finding.

79

Maladaptive Protection Motivations-Gain Condition
When self-efficacy and locus of control increase, maladaptive protection motivations
decrease. Self-efficacy showed a much stronger, negative relationship to maladaptive behavior
than internal locus of control. Consumers must be willing and able to avoid potential losses
inferred by maladaptive responses to ensure they achieve similar gains personally. A consumer’s
willingness and capacity to avoid losses is dependent on a perceived benefit of not taking action,
such as saving time, that a maladaptive response offers. This finding suggests that consumers
feel a strong desire to react to the recall message when faced with potential losses regardless of
their willingness and ability to react. A similar finding in the gain frame suggests that selfefficacy and internal locus of control work together to prevent injury through two key processes.
First, self-efficacy decreases the potential for maladaptive protection motivations. Second, as
participants perceive that preventing the recall risks is within their control, they are more likely
to seek the implied gains in the message. Prospect theory suggests that people prefer gains over
losses and when presented with a gain oriented message (Tversky & Kahneman, 1981), they are
averse to risks and lessen their potential for maladaptive responses. These findings support prior
PMT research with greater influence of coping and efficacy on protection motivations (Rippetoe
& Rogers, 1987; Boer & Seydel, 1996; Milne et al., 2006).

Maladaptive Protection Motivations—Control Condition
In the control condition, when severity and self-efficacy increase, maladaptive protection
motivations decrease. Self-efficacy demonstrated a strong inverse relationship to maladaptive
protection motivations. Severity perceptions were only significantly related to maladaptive
protection motivations in the control condition. This finding suggests that participants evaluate

80

severity differently when no framing is present, and the elicited fear increases severity
perceptions. The absence of statements of expected gains and losses may explain severity
perceptions in the control condition. The recall stimulus in the control condition lacked the
specific information about potential illness present in the gain and loss conditions. The lack of
information about potential illness in the gain and loss conditions may explain differing
perceptions of the severity of the recall message (Maheswaran, 1990; Block & Keller, 1995). It
may also show that lack of specific information about potential illness results in the control
message elicited doubt and thus influenced perceived severity. Whether recall messages should
include specifics about potential for injury or provide other useful information to reduce
maladaptive responses needs further research. We view this discovery as promising for recall
message research.

Summary of Maladaptive Protection Motivations
The loss message frame sparked more concern about potential injury among subjects of
than either the gain or control conditions. Stronger negative weights of self-efficacy on
maladaptive protection motivations in the loss condition compared to the gain and control
conditions, indicates that consumers have greater motivations to avoid potential illness. This
finding agrees with Meyerowitz and Chaiken’s (1987) prospect theory findings that loss frames
lead to higher persuasiveness for breast self-examination and that loss frame participants were
more likely to seek mammography (Banks et al, 1995).
Our results show that consumers do not weigh coping responses when considering
maladaptive responses (not reacting to the recall message). Coping with recall risks underlies
the challenges of creating ongoing behavior change and consumer engagement in prevention

81

behaviors. The potential of locus of control to reduce maladaptive reactions to risks supports
prior research of prevention strategies relationship to gain frames in prospect theory (Deitweiler
et al, 1999). The probability of the lower risk of illness (gain frame) may persuade participants
that it is within their control to prevent the risks from the recall based on the message text they re
Manipulation of gain oriented messages may produce the dual effect of reducing
maladaptive responses while increasing adaptive responses due to the negative effects of selfefficacy and perceived locus of control on maladaptive protection motivations. Loss framed
messages also showed a negative relationship of self-efficacy on maladaptive protection
motivations. Loss oriented messages may increase a consumer’s willingness to take action to
avoid a negative outcome, or facilitate risk aversion, thus supporting gain seeking implied in
prospect theory.

Adaptive Protection Motivations—Loss Condition
We found that when pro-active coping, reactive coping, locus of control and confidence
in the communicator increase, adaptive protective motivations increase.

When participants read

that 20% of subjects will become seriously ill, their personal ability to deal with risks increases
and they are more likely to adopt adaptive protection motivations. Participants likely feel that
the combination of a trusted communicator and their ability to react by adopting a coping
strategy will reduce the probability of illness.
Coping plays a role in adaptive protection motivations not exhibited in maladaptive
protection motivations in the loss condition, and suggests a greater impact of loss framing on
eliciting adaptive protection motivations than gain messages or the control condition. This

82

finding supports the findings of Meyerowitz & Chaiken (1987); Detweiler et al., (1999) who
demonstrated the relationship between loss framed messages and adaptive responses.
In the loss message frame, confidence in the communicator also displayed a positive
influence on adaptive protective motivations. Consumers feel that they must be confident in the
source of the recall message to engage in adaptive protection motivations, in this case, a retailer.
The role of a credible communicator when subjects face risk appears to demonstrate a strong
influence on protection motivations. Understanding the impact of gain or loss frames on
perceptions of message communicator credibility, requires further research. When facing sure
losses (20% seriously ill), if a consumer believes that he has the ability to deal with the situation
(locus of control) and is ready to take action (coping) and the message is from a trusted
communicator, then he will likely take adaptive steps to prevent illness. Discovering the
influence of trusted communicators in recall communications provides new knowledge for
prospect theory and protection motivation research as an important variable of interest.

Adaptive Protection Motivations—Gain Condition
We found that when self-efficacy, proactive coping, reactive coping and confidence in
message communicator increases, adaptive protection motivations increase. The strongest
relationship for the gain condition to adaptive protection motivations was reactive coping. This
finding is logical because people are likely to perceive greater rewards and lessen potential losses
if they are willing and able to react to the recall message.
Gain framed messages showed a different effect than loss frames by a stronger influence
between self-efficacy to engage in adaptive protection motivations. The assurance that 80% of
people will not become ill strongly influenced the willingness and ability of participants to take

83

adaptive protection behaviors. This finding is likely because of the higher implied probability of
80% avoiding injury is perceived by participants as a gain, thus increasing peoples’ willingness
and ability to act on the recall message. Reactive coping showed greater influence on adaptive
protection motivations than proactive coping.

Reactive coping implies gain seeking behavior to

achieve similar personal outcomes by participants and supports prior research where gain frames
demonstrated greater influence on protection motivations than loss frames. (Detweiler et al.,
1999; Rothman & Salovey, 1997). Locus of control, not previously studied in framing
manipulations, appears to influence participant’s perceptions differently in gain frames, and
moderates adaptive protection motivations. This finding supported O’Keefe & Jensen (2006) and
Block & Keller’s (1995) studies that adaptive and maladaptive responses are complicated by
other cognitions than classical gain and loss frame manipulations elicit. The cognitions are
likely predicated on self and response efficacy variables which mediate protection motivations
through the coping response, as opposed to a direct relationship to protection motivations as
modeled in this study. Since self-efficacy significantly influenced adaptive protection
motivations in the gain frame and maladaptive protection motivations in the loss frame, a test of
self-efficacy mediation through the coping variable should be modeled. Fear created from the
loss message may be a better motivator of preventive action than the gain message (80% of
people not getting ill).

Adaptive Protection Motivations—Control Condition
We found that when self-efficacy, locus of control, proactive and reactive coping and
confidence in message communicator increase, adaptive protection motivations increase.
Clearly, severity is perceived differently in the control message than in gain/loss frames. This

84

phenomenon requires further study about the role of threat perceptions when probabilistic
outcomes are presented to people. Similar to the maladaptive protection motivation findings,
self-efficacy, displayed a role in adaptive protection motivations in both the gain and control
conditions, but not in the loss conditions. Confidence in the communicator demonstrated the
strongest relationship to adaptive protection motivations in the control condition. Without
framing, confidence seems more important in motivating adaptive protection motivations.

The

relationship of locus of control and coping may be similar to self-efficacy and coping in ordered
protection motivation theory, requiring additional modeling (Rippetoe & Rogers, 1987).

Summary of Adaptive Protection Motivations
Coping appraisals in the loss condition demonstrated a strong positive relationship to
adaptive protection motivations and this finding provides support to the controversy in the PMT
literature about the value of loss framed messages influence on coping behaviors and subsequent
protection motivations compared with gain framed messages (Keller and Block, 1996; Rogers,
1983; Tanner, et al., 1991; Witte, 1994). We found similar strong relationships for loss messages
on adaptive protection motivations. Clearly, framing the message in gain and loss formats
influences peoples’ protection motivation, whether they are adaptive or maladaptive in nature;
however, further research is required to produce more precise message manipulations.
Self-efficacy also appears to be an important influence on adaptive protection
motivations as shown in both the gain and control messages and insignificant role in the loss
message. The fact that self-efficacy was not significant in the loss condition is surprising. Selfefficacy has been demonstrated as an important factor in loss framed message manipulations
influence on protection motivations (Keller and Block, 1996; Tanner et al., 1991). It was also

85

unexpected that locus of control would be significant in the loss condition, but not in the gain
condition on adaptive protection motivations, but significant on maladaptive protection
motivations. Locus of control may be less influential when people are weighing potential gains.
The finding that self-efficacy was not significant in the loss frame on adaptive protection
motivations may signal that the fear appeal is less effective because of the stated probability of
illness in the message, as opposed to the higher probability of people being able to avoid illness
in the gain condition. Locus of control is likely a precursor to peoples’ assessment of selfefficacy by an understanding they can control the risk through their actions.

Since the weight

of self-efficacy on adaptive protection motivation was higher in the gain condition compared
with the control condition, we propose that self-efficacy plays an important role when subjects
are presented probabilistic data like the percentages of people who will become ill. Positive
framing may lessen fear, which in turn, promotes desirable actions.
Confidence in the message communicator was another dimension we measured with
surprising results. Confidence in a message communicator does not influence maladaptive
protection motivations in any of the three conditions in the observed negative relationships of
other variables tested. While this relationship has not been tested in the product recall literature,
the finding supports Jones et al. (2003) exercise intentions research. This study demonstrated that
only positive messages from a credible source were effective in influencing adaptive behavior
change. These findings also support Arora’s (2006) findings that credible communicators have
greater influence on adaptive behaviors, than on maladaptive behaviors. The unanticipated nonsignificance for confidence in the message communicator and maladaptive protection
motivations may be due to cognitive dissonance that participants trust the communicating party,
but have no intention of following their direction for many reasons related to cost of response

86

weighting. Since the positive sign existed for all three experimental conditions, the role of the
message communicator to influence positively desirable or undesirable responses to recalls will
need further study.
The contrasting role of proactive and reactive coping on adaptive protection motivations
in message frames is the most interesting finding in this study. Coping in loss framed messages
increases adaptive protection motivations. Coping in gain framed messages demonstrates a
weaker influence on adaptive protection motivations than in loss framed messages, and suggests
that loss frames offering probabilities of people getting ill are more effective at influencing
protective behavior. Probabilities of persons who will not become ill presented in message
content appear to be less effective for influencing protective behavior.

Implications
Stakeholders in the consumer goods market involving recalls are served some
important food for thought from this study. Reinforcing trustworthiness, skill, and confidence by
those agents communicating recall messages to consumers have a strong relationship to
consumers’ adaptive protection motivations. Message framing may increase the influence of key
factors which are likely to increase protection motivations. For messages which use gain
oriented message content, key dimensions influencing protection motivations are self-efficacy,
confidence in the message communicator and coping responses. For loss oriented messages,
self-efficacy, confidence in the message communicator, coping responses, and the added
dimension of locus of control show greater influence on adaptive protection motivations. For
control recall messages, self-efficacy, confidence, and coping were also influential, although less
so than in the loss condition.

87

Susceptibility and severity of risks did not influence consumers’ protection motivations
in framed messages. As discussed earlier, the self-positivity biases and discounting effects in the
face of real risks from defective food products are largely ignored by customers. Peoples’
strategy for coping where food recalls are routine seems to be linked to specific experiences in
the past that shape current and future behaviors. Stated probabilities of illness appear to bring the
message more relevance to people reading recall messages.
Our research of recall messages has continued for over four years. During this time we
have formed some insights about evolving ways that retailers have tried to communicate more
effectively with customers and help them prevent injuries from recalled products. We studied
low-tech mom and pop retailers who use only an information binder at their service desk to track
current recalls or post recall notices on a bulletin board full of competing marketing and
informational messages. We studied regional mass merchants harnessing the full power of
analytics and customer relationship management data mining to contact each customer who
bought a recalled product. This continuum of recall communication still faces problems as “big
brother” analytic methods to harness and manipulate data for targeted consumer behaviors
increase exponentially in the retail sector. Retailers continue to compete for consumer attention
through new technologies by understanding and using personal information to facilitate
marketing.
One consistent observation that led to this study was attempts by retailers to deploy novel
technologies (e.g. email, cellular, Internet, online advocacy groups) to address the recall
communication problem. Informing customers about recalls and creating appropriate responses
to recall risks remains a challenge in spite of new technologies. Just a decade ago, recall
messages were communicated through two primary sources: signage or printed messages in

88

stores and the news media wires. Recall messages were delivered to fewer outlets than current
communication channels allow and many consumers were never informed about the risks. More
consumers hear about recalls today through ever increasing media outlets, but the response rate
remains abysmal. A hit or miss approach to informing consumers about recalls is inadequate for
many reasons. Recalls may damage brands, corporate reputations, financial performance, raise
health care costs for industry and consumers, and most importantly, threaten human health.
In the past decade, retailers have attempted to use various communication innovations
including: loyalty data notifications, receipt notification, automated telephone notifications by
zip code, mass letter writing campaigns, subscription email notifications, subscription cellular
texts, Twitter feeds, consumer advocacy sites posting recall data, government websites, retailer
traceability improvements, supply chain improvements and retailer sites with recall notice
sections. Despite these uncoordinated tries at creating a compelling innovation to perfect recall
message communication, we have learned that recall messages are still not getting through to
information overloaded consumers. Indeed, recall message senders must champion their
formative role to influence consumers by first prompting detection of food borne risks and then
to help prevent them with adaptive consumer coping. On one hand, the recall message content
must increase perceptions of sure gains, but on the other hand decrease perception of sure losses.
This may be possible in one recall message by creating parallel constructions using opposite
statements of probability throughout the message content. For example, a recall message
detailing the percentages of people in high risk groups that will become ill and percentages of
low risk groups that will not become ill would be a parallel construction.
Our findings suggest that rather than relying on dramatic new technological innovations
to communicate recalls, the focus should be on the content of the recall message. We must

89

understand the influence of message frames on consumers that may motivate socially oriented
behavior change. This may be achieved through what Snow and Benford (1988) have deemed
“frame alignment”. Frame alignment occurs when frames are balanced and complementary,
creating a “resonance” and allowing predictable subject responses to message content
manipulations. An example is including two suitable messages, each framed differently, in the
same recall notice such as percentages of people who will and will not become ill to find out if
consumers report different protection motivations than in separate frames (i.e. gain or loss).
Frame alignment manipulations may influence desirable behavioral results when consumers face
risk. Consumers who understand their personal potential for illness are both better informed and
able to protect themselves. The goal of frame alignment was displayed in our study when
subjects showed negative relationships to maladaptive protection motivations and positive
relationships to adaptive protection motivations. Our goal in recall message design is to decrease
maladaptive responses and increase adaptive responses and to this extent we were successful. We
need to understand how to include different levels of probabilistic statements to increase
adaptive protection motivations. The relationships of message framing on protection motivations
is less predictable than hypothesized and indicates loss based messages are more effective for
influencing protection motivations. We need a better understanding of the value of gain and loss
framing to motivate desirable behaviors.
Most people, based on experience, think that events in their life will go according to their
expectations. Sadly, many people underestimate risk potentials and face unnecessary
consequences, especially for food borne illness. For this study, we found that threat assessments
were not predictors of protection motivations in any of the three conditions (control, gain, loss).
This suggests that covariates such as trust, blame, locus of control, experience and more

90

sophisticated cognitive processes are involved in eliciting an immediate reaction than simply
presenting a threat message. Some threats consumers face are of immediate concern while
others are background noise and may be discounted. As Kahneman and Tversky (1979) have
suggested, consumers must apply complex weighting schemes to prioritize which threats have
meaning. Recalls are low priority for most of us because we haven’t experienced life threatening
illness, and we have a preference to hope for the best result when faced with risks. Ultimately,
consumers may need to experience significant feelings of loss for us to develop feelings of deep
regret and stimulate an epiphany for changing our behaviors permanently to avoid future risks.
We are only beginning to understand those causes that influence protection motivations
favorably, but this study has made headway in our search for knowledge.
Consumer notification is one of the most challenging processes in product recall
management. Products may be in the consumer’s home and present a race against time to
prevent imminent illness by the consumer’s failure to lessen or erase the risks. Because of the
critical risks to consumers from food recalls, prospect theory may be an effective tool to help
frame recall messages that elicit favorable and timely responses by stimulating predictable
behaviors in consumers. In keeping with prospect theory, loss framed messages stimulate
consumer fear and produce two desirable effects that offer promise for future research.
Another dimension in creating adaptive responses to recall messages involves the
perceived credibility of the message communicator. Perceived credibility influences adaptive
responses but demonstrated no significant relationship to maladaptive responses. The role of
credibility indicates that the consumer who reads a framed recall communication takes the
message and its communicator seriously and plans to react. The importance to the consumer of a
credible communicator suggests the consumer realizes a retailer is an important reference group.

91

Finally, this recall message framing study offers consumers likely results if they take
action. The message frames offer some assurance of illness or avoiding illness. Consumers are
more averse to losses than potential gains when coping with a recall threat as evidenced by our
2
R results. Framing appears to influence a consumer’s perception that a desirable outcome is

within their control and manifests through a greater willingness and ability to personally act to
reduce these implied risks.

For practitioners, this study suggests loss framing will influence

protection motivations more than gain and control framing.

Recommendations for Further Research
We have discovered several opportunities for future research based on this study:
1. We plan to design future experiments which manipulate different types of message
content to better understand those messages that influence adaptive behaviors.
2. We plan to add additional variables to our model with the goal of explaining additional
variance within protection motivations including: experience with recalls (impact), risk
level (low health risk/high health risk, blame attributions, social responsibility of the
message communicator, and others.
3. We will target broader samples of individuals to participate in our studies to increase
validity by surveying underrepresented groups in the United States, such as minorities,
elderly persons. immune system compromised individuals, and varied socioeconomic
groups who might be less prone to having access to communication innovations for
communicating recalls.
To achieve frame alignment as suggested by Snow & Benford (1988), the manipulation
of gain and loss content must be adjusted to determine if a parallel relationship of the variables to
92

adaptive and maladaptive protection motivations exists. This may be achieved through a more
comprehensive analysis of the impact of message heuristics perceptions by subjects at a higher
level of precision than this study allowed. Key research areas we will focus on include:
1. Recall problem definitions
2. Attributions of blame called for by product failures requiring what consumers consider
non normal proper handling tactics (e.g. cooking thermometers) to avoid risks from the
recalled product.
3. Useful information framing suggesting proper handling tactics and other solutions to
avoid risk that are actionable for most consumers.
4. Motivational framing such as avoiding serious injury frames in this experiment implying
socially accountable consumer actions (e.g. protect your family and friends).
5. Manipulations of relevance to the individual consumer such as manipulating
measurements for those who have been or could be impacted by the recall framing.

We started our journey for this study by examining how a major retailer communicates
recalls to their customers. The alerts messages used as a control (Figure 3b) in our study is a
typical message in the retail Internet grocery sites to alert customers. The message is
problematic in many ways when compared to the FDA press release for the same recall in Figure
3a, but we kept the retailer’s content for this experiment to gain an understanding of a typical
message’ effectiveness on consumers. We believe that recalls are central to only a small number
of consumers’ day-to-day threat assessments and decision making. For the rest, the solution to
improving responses to the pervasive problem of tainted food is not so simple. We may not feel
motivated to audit retailer websites on a regular basis to protect our health. Clearly, this study

93

showed the importance of the consumer’s locus of control when facing gain and loss scenarios.
Consumers in the United States clearly want recalls to be within their power to react (locus of
control) and eliminate the risks, as opposed to relying on the government or producers to erase
risks. On the other hand, if consumers believe that they are personally accountable for food
borne illness from improper preparation, they might have different perceptions of locus of
control and ways to cope with current and future risks. The relationship of variables such as
locus of control and experience will be studied in the future.
All food involves some measure of risk to consumers, but often these risks are under the
consumer’s control. When foodborne pathogens are discovered, retailers must take responsibility
for their errors and help lessen immediate risks to customers. Responsibility to customers may
be supported through better detection and prevention information offered in recall messages.
Detection message content may describe ways to identify if a product is tainted by using cues to
identify the problem (e.g. coding, packaging). Prevention message content may describe proper
preparation and storage to avoid potential illness.
Based on some of the protection motivation variable loadings found, additional
measurement model development is necessary. Some areas for change include: susceptibility,
severity, self-efficacy, and potentially others as well. The susceptibility to and severity of the
threat of recalled products may need to be expressed more dramatically to influence subjects.
Protection motivation classifies risk communications in two dimensions (susceptibility and
severity) but these dimensions have done a relatively poor job of explaining consumer
perceptions. While Keller and Block (1996; 1998) had some success in manipulating severity in
preventative actions to improve protection motivations, other research has been mixed on threat
messages and consumer threat assessments. One possible reason for this is when we consider the

94

typically unmeasured variable of experience with a threat as a mediating state, we are likely to
encounter different perceptions of severity of the potential risks and susceptibility to them for
individuals assessing risk. Future research should compare respondents’ interpretations of
susceptibility and severity to assess the influence of experience (e.g. diagnosed with food borne
illness) on susceptibility and severity perceptions among participants and what role these
experiences have on protection motivations.
Likewise, the message treatment stimuli in future studies should be more inclusive of all
potential consumers compared to the control message which identified pregnant women and
people with compromised immune systems. Other groups such as the elderly and small children
have high susceptibility to food borne illness and should be added to messages to increase
generalizability. These goals will be achieved through refining survey items, stimuli, and
executing a similar study in a laboratory setting. This laboratory experiment will accomplish
several goals. It will allow us to test more effective scales for measuring the theories and
provide information about the differences between laboratory and online research methods.
A modeling technique employing statistical weights to estimate statistical norms in the
United States population should be tested to discover what differences exist, if any, between the
sampling distributions in this study.

Limitations of the Study
Convenience samples are limited by geodemographic inequivalence to the United States
population. Participants taking surveys for financial incentives may possess resources that other
potential subjects of interest do not. For example, access to the Internet, and qualification to take
part in the sample by membership in MTurk are potential limitations of this sample. The sample

95

possesses higher levels of education than the general population. Less educated individuals, who
could be more prone to food borne illness, were underrepresented in this study and decrease the
generalizability of the findings.
A single item stimulus involving fruit may not be reflective of poorer families essential
shopping cart contents. For this reason, a basket of top priority goods for consumers that might
be prone to recalls may provide different insights than contaminated fruit being tested as the
stimulus in this experiment.
Protection motivation theory and prospect theory provide interesting insights into the way
the world works. Combining theories with different findings does not seamlessly provide a
model for interpreting the reality that underlies social science data. Statistical assumptions for
maximum likelihood estimation may not conform to the nuances of data, measurement, and
causal relationships. However, theoretical bases provide an excellent starting point for
attempting to understand consumer behavior from risks of defective food products. This study
clearly showed that wording and measurement issues in social science research are challenging
to fit suitably to a diverse sample and each study offers clues for developing a better research
agenda in the future. Qualitative mixed method research methods would have benefited this
study by helping to develop more effective survey items. Modeling of data to typical populations
in the country studied through weighting techniques should increase the generalizability of the
findings. In our study Asians were disproportionately represented compared to the U.S. census
population and Hispanics were disproportionately underrepresented in the study. By applying
data weights to each racial group to census estimates, our findings may be more generalizable to
the U.S. population.

96

Designing a study which is relevant for diverse groups of consumers using hypothetical
information is less valuable than live data in real world recall events. A test of the impact on U.S.
citizens during a real recall event and the impact of this real-time event on protection motivations
could be more effective than an experimental design.

97

APPENDICES

98

APPENDIX A
TABLES

Table 1. Preliminary Survey Demographics
Age 18-24 Convenience
Sample
Male
Female
Total Valid

Traditional Grocery
Mass Retailer
Club
Local Independent
Total Valid

Dairy
Bakery
Fruits
Vegetables
Eggs
Salty Snacks
Meat/Poultry/Fish
Rice/Pasta/Pizza
Peanut Butter
Cold cuts/deli
meat/hotdogs
Ethnic Foods
Soups/Chili
Spreads/Dressings
Sweets/Desserts
Total Valid

Gender
32
46
78

41%
59%
100%

Store type buy most groceries
23
51
3
1
78

29%
65%
4%
1%
100%

Shopping Basket Composition
(Monthly)
61
59
58
58
49
48
45
45
36

78%
76%
74%
74%
63%
62%
58%
58%
46%

34
34
31
30
27
78

44%
44%
40%
38%
35%
100%

99

Table 1 (cont’d).

Lettuce
Fruit
Deli Meats
Ground Beef
Cheese
Ground Turkey
Total Valid

Top Basket Priority (Typically
Recalled Products)
5
32
19
15
5
2
78

6%
41%
24%
19%
6%
3%
100%

Store Brands
National Brands
Total Valid

Brand Preference
41
34
75

55%
45%
100%

Yes
No

Kroger Familiarity
67
8

89%
11%

100

Table 2. Survey Instruments

Dimension

Manipulation
Check
Manipulation
Check

Manipulation
Check

Manipulation
Check

Item
The recall message
emphasized what
might happen in the
future if I ate the
cantaloupe.
The recall message
emphasized things
that can happen
immediately if I ate
the cantaloupe.
The recall message
emphasized the
health benefits of
not eating the
cantaloupe.

Adapted from

The recall message
emphasized the
risks of eating the
cantaloupe.

The message
emphasized things that
can happen in the
future, Meyers (2010)
The message
emphasized things that
can happen
immediately, Meyers
(2010).

Cronbach's
Alpha
Scale

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

The message
emphasized the benefits
of being physically
active, Meyers (2010).
N/A

1=Strongly
disagree
7=Strongly
Agree

The message
emphasized the risks of
not being physically
active, Meyers (2010).

1=Strongly
disagree
7=Strongly
Agree

101

N/A

Table 2 (cont’d).
Threat
AssessmentSusceptibility
Threat
AssessmentSusceptibility

Threat
AssessmentSusceptibility
Threat
AssessmentSusceptibility

Threat
AssessmentSeverity

Threat
AssessmentSeverity

Threat
AssessmentSeverity

Threat
AssessmentSeverity

I am at risk of
getting sick if I ate
this cantaloupe.
I believe that
protecting myself
from food borne
illness by the
information
presented in this
cantaloupe recall
message is

Threats to my health
if I bought and ate
this cantaloupe are

The bacteria
discussed in this
cantaloupe recall
message is
It could seriously
injure me if I didn't
react to the
instructions in this
recall message and
ate the cantaloupe.

N/A

How severe are the
consequences of the
disease, Boer & Seydel
(1996).
I think breast cancer is a
more serious disease
than other diseases I
know, Boer & Seydel
(1996).
Despite the advances of
medical science breast
cancers remains as
serious as it was in
former days, Boer &
Seydel (1996).

N/A

102

.71
(scale)

1=Highly
Unlikely
7=Highly
Likely
1=Strongly
disagree
7=Strongly
Agree

N/A

1=not at all
important
7=very
important

.71
(scale)

1=Not at all
Severe
7=Very
Severe

.71
(scale)

1=Harmless
7=Very
Severe

N/A

How severe are the
consequences of the
disease? Rogers (1983).
It is rather probable that
I will ever get breast
I will get sick if I eat cancer, Boer & Seydel
this cantaloupe.
(1996).
The chance that
someone of my age in
The likelihood of me comparable conditions
getting sick from
gets breast cancer is
eating this
rather large, Boer &
cantaloupe is.
Seydel (1996).

.71
(scale)

N/A

I am vulnerable to
food borne illness if
I eat this cantaloupe.

1=not at all
vulnerable
7=very
vulnerable
1=Strongly
disagree
7=Strongly
Agree

1=Strongly
disagree
7=Strongly
Agree

N/A

Table 2 (cont’d).

Locus of
Control

Locus of
Control

Locus of
Control

Locus of
Control

Protecting myself
from injury based on
in information in
this cantaloupe
recall message is
It is within my
control to protect
myself based on the
information in this
cantaloupe recall
message.
The primary
responsibility for
protecting myself
from food borne
illness presented in
this cantaloupe
recall message
belongs to MYSELF
The primary
responsibility for
protecting myself
from food borne
illness presented in
this cantaloupe
recall message
belongs to OTHERS

Keeping my
confidential information
safe is, (Workman et al.,
(2008).
I believe that it is within
my control to protect
myself from
information security
violations, (Workman et
al., (2008).

The primary
responsibility for
protecting my
confidential information
belongs to, (Workman
et al., (2008).

The primary
responsibility for
protecting my
confidential information
belongs to, (Workman
et al., (2008).

103

.88
(scale)

1=completely
beyond my
control
7=completely
within my
control

.88
(scale)

1=Strongly
disagree
7=Strongly
Agree

.88(scale)

1=Strongly
disagree
7=Strongly
Agree

.88(scale)

1=Strongly
disagree
7=Strongly
Agree

Table 2 (cont’d).

Response
Efficacy

Using the
information in this
cantaloupe recall
message will help
me prevent injury.

N/A

N/A

Response
Efficacy

Responding to this
cantaloupe recall
message is effective
for avoiding injury? N/A
This cantaloupe
recall message
offered valuable
information for ways
to reduce the risk of
injury.
N/A

Self-efficacy

I feel insecure in my
ability to respond to
this cantaloupe
recall message.

Response
Efficacy

Self-efficacy

Self-efficacy

Self-efficacy

Self-efficacy

I would avoid taking
action to this
cantaloupe recall
message.
Responding to
this cantaloupe
recall message is too
complicated and I
will not even bother
to try.
I am able to use
the information in
this cantaloupe
recall message to
prevent injury.
I would give up
on responding to this
cantaloupe recall
message easily

1=Not at all
effective
7=Very
Effective

N/A

N/A

1=Strongly
disagree
7=Strongly
Agree

1=Strongly
disagree
7=Strongly
Agree

I feel insecure about my
ability to do things,
(Reversed) Sherer et al.
(1982).

1=Strongly
disagree
7=Strongly
.55 Factor Agree

I avoid facing
difficulties, (Reversed),
Sherer et al. (1982).

1=Strongly
disagree
7=Strongly
.67 Factor Agree

If something looks too
complicated, I will not
even bother to try it,
(Reversed) Sherer et al.
(1982).
When unexpected
problems occur, I don't
handle them well,
(Reversed) Sherer et al.
(1982).

I give up easily, Sherer
et al. (1982).

104

1=Strongly
disagree
7=Strongly
.68 Factor Agree
1=Strongly
disagree
7=Strongly
.55 Factor Agree
1=Strongly
disagree
7=Strongly
.69 Factor Agree

Table 2 (cont’d).

Coping
Response

Coping
Response

Coping
Response

Coping
Response

Coping
Response

Coping
Response

Coping
Response

I will make a plan
of action to protect
myself when I read
a recall message
like this one that
may impact me or
my family.
I will make a plan
of action when I
read a recall
message like this
one to protect
myself.
I will create
solutions to avoid
food borne illness
from recalled food
products like the
cantaloupe.
I will think about
the best way to
handle the risks
from the
cantaloupe and
avoid injury.
I will concentrate
my efforts on doing
what the
cantaloupe recall
message suggests
to avoid injury
from recalled
products.
I will do what has
to be done
according to this
cantaloupe recall
message to avoid
injury.
I will follow the
information in this
cantaloupe recall
message to avoid
injury.

Concentrate on ways
the problem could be
solved, Duhachek
(2005).

Try to make a plan of
action, Duhachek
(2005).

Generate potential
solutions, Duhachek
(2005).

Think about the best
way to handle things,
Duhachek (2005).

Concentrate my efforts
on doing something
about it, Duhachek
(2005).

Do what has to be
done, Duhachek
(2005).

Follow a plan to make
things better,
Duhachek (2005).
105

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

.87 Scale

1=Strongly
disagree
7=Strongly
Agree

Table 2. (cont’d).

Protection
Motivation-Maladaptive
Coping
Protection
Motivation-Maladaptive
Coping

Protection
Motivation-Maladaptive
Coping

I do not need to
respond to this
cantaloupe recall
message because I
am protected by my
genes.
I believe that I am
lucky enough to
avoid illness from
this cantaloupe
recall.
I am unlikely to get
sick if I eat the
cantaloupe in this
recall message
because I have
never had any
serious health
problems in the
past.

I do not need to engage
in health behaviors
because it one's genes
and not one's behaviors
that determine whether
one is healthy or not,
Ho et al. (2005).

Staying healthy is a
matter of luck, Ho et
al. (2005).

I am unlikely to get
seriously ill as I have
never had any serious
health problems in the
past, Ho et al. (2005).

106

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

.67 Scale

1=Strongly
disagree
7=Strongly
Agree

Table 2 (cont’d).

Protection
MotivationAdaptive
Coping

Failure to engage in
specific health
behaviors like
I am better able to
nutritious eating,
protect myself and exercise, regular
my family because medical check-ups,
of the information
etc. will increase the
presented in this
risk of developing a
cantaloupe recall
preventable disease,
message.
Ho et al. (2005).
Avoiding regular
exercise and physical
Ignoring recall
activity will increase
messages in the
the risk of developing
future will increase health problems, Ho et
my risk of illness
al. (2005).
I am confident that I
can make healthy
I am confident I
lifestyle changes (e.g.
will use recall
diet, exercise, reduce
messages in the
tobacco/alcohol intake,
future to protect my etc. if I had to, Ho et
health.
al. (2005).

Protection
MotivationAdaptive
Coping

I will use recall
message information
in the future to
reduce the risk of
serious illness from
eating defective food
products.

Protection
MotivationAdaptive
Coping

Conducting regular
self-examinations
(breast or testicular) is
an effective means of
Using information
reducing the risk of
in recall messages
serious/health
is an effective
problems, including
means to reduce the breast or testicular
risk of serious
cancer, Ho et al.
illness.
(2005).

Protection
MotivationAdaptive
Coping

Protection
MotivationAdaptive
Coping

Having regular medical
check-ups is effective in
reducing the risk of
serious health problems,
including breast and
testicular cancer, Ho et
al. (2005).

107

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

N/A

1=Strongly
disagree
7=Strongly
Agree

Table 2 (cont’d).
Kroger has a great
amount of
experience to create
this cantaloupe
recall message.

XYZ Corporation has a
great amount of
experience, Newell &
Goldsmith (2001).
.84-.92

1=Strongly
disagree
7=Strongly
Agree

XYZ Corporation is
skilled at what they do,
Newell & Goldsmith
(2001).

.84-.92

XYZ Corporation has
great expertise, Newell
& Goldsmith (2001).

.84-.92

1=Strongly
disagree
7=Strongly
Agree
1=Strongly
disagree
7=Strongly
Agree

Perceived
Confidence of
the Message

Kroger has the
skills to create this
cantaloupe recall
message.
Kroger has great
experience to create
this cantaloupe
recall message.
Kroger does not
have much
experience to create
this cantaloupe
recall message.

.84-.92

1=Strongly
disagree
7=Strongly
Agree

Perceived
Confidence of
the Message

Kroger is making
truthful claims in
this cantaloupe
recall message.

The XYZ Corporation
makes truthful claims,
Newell & Goldsmith
(2001).

.84-.92

1=Strongly
disagree
7=Strongly
Agree

This cantaloupe
recall message from
Kroger is honest.
I do not believe
what Kroger is
telling me in this
cantaloupe recall
message.
I trust this
cantaloupe recall
message from
Kroger.

The XYZ Corporation
is honest, Newell &
Goldsmith (2001).

.84-.92

1=Strongly
disagree
7=Strongly
Agree

Perceived
Confidence of
the Message
Perceived
Confidence of
the Message
Perceived
Confidence of
the Message

Perceived
Confidence of
the Message

Perceived
Confidence of
the Message
Perceived
Confidence of
the Message

XYZ Corporation does
not have much
experience, Newell &
Goldsmith (2001).

I do not believe what
the XYZ Corporation
tells me, Newell &
Goldsmith (2001).

.84-.92

I trust the XYZ
Corporation, Newell &
Goldsmith (2001).

.84-.92

108

1=Strongly
disagree
7=Strongly
Agree
1=Strongly
disagree
7=Strongly
Agree

Table 3. Demographics
Combined
Samples
Frequency
Gender
Male
431
Female
459
Total
890
Age Groups
18-24 years
147
25-34 years
357
35-44 years
144
45-54 years
134
55-64 years
85
65 years +
22
Total
889
Marital Status
Single
375
Married
449
Separated
6
Divorced
48
Widowed
8
Total
886
Race
White
468
African
American
41
Hispanic
22
Asian
302
Native
American
19
Pacific Islander
3
Other
32
Total
887

Percent
48.4
51.6
100
16.5
40.2
16.2
15
9.6
2.5
100
42.3
50.7
0.7
5.4
0.9
100
52.8
4.6
2.6
34
2.1
0.3
3.6
100

109

Table 3 (cont’d).
Family
Structure
In a marriedcouple family
In a family with
female
householder, no
spouse present
In a family with
male
householder, no
spouse present
In a group of
unrelated
subfamilies
Unrelated
individuals
Total
Recall
experience
Have you
purchased a
food item that
was later
recalled
Yes
No
Total
Have you ever
been diagnosed
with food borne
illness by a
medical
professional
Yes
No
Total

539

61.3

126

14.3

76

8.6

30

3.4

109
880

12.4
100

369
517
886

41.6
58.4
100

219
666
885

24.7
75.3
100

110

Table 3 (cont’d).
Current Occupation
Managerial or
professional
Executive,
administrative

Frequency

Percent
88

10.1

86

9.9

160

18.4

45

5.2

60
73
85
72
110
6

6.9
8.4
9.8
8.3
12.6
0.7

23

2.6

Farming, forestry, or
fishing

9

1

Precision production,
craft, or repair

10

1.1

Operations, fabrication,
or labor

11

1.3

13

1.5

12

1.4

7
870

0.8
100

Professional specialty
Technical, Sales, admin
support
Technicians or related
support
Sales
Administrative
Service
Private household
Protective service
Service exc. Protective
or household

Machine Operation,
assembly,
Transportation or
material moving
Handling or cleaning
equipment
Total

111

Table 3 (cont’d).
Education
Less than High
School
High
School/GED
Some College
2 Year College
Degree
4 Year College
Degree
Master’s
Degree
Doctoral
Professional
(JD, MD)
Total
Annual
Income
Below $20,000
$20,000$29,999
$30,000$39,999
$40,000$49,999
$50,000$59,999
$60,000$69,999
$70,000$79,999
$80,000$89,999
$90,000 or
more
Total

5

0.6

63
168

7.1
19

85

9.6

342

38.6

192
14

21.6
1.6

17

1.9

886

100

241

30.9

166

20.5

112

13.2

110

12.3

77

8.5

43

4.7

42

4.2

21

1.8

62
874

3.9
100

112

Table 3 (cont’d).
SAMPLE 1
MTurk
Frequency Percent
Gender
Male
401
52%
Female
370
48%
Total
771
100%
Age Groups
18-24 years
145
18.8%
25-34 years
339
44%
35-44 years
119
15.5%
45-54 years
96
12.5%
55-64 years
57
7.4%
65 years +
14
1.8%
Total
770
100%
Marital Status
Single
361
46.9%
Married
366
47.6%
Separated
5
0.7%
Divorced
31
4%
Widowed
6
0.8%
Total
769
100%
Race
White
362
47%
African American
39
5.1%
Hispanic
21
2.7%
Asian
299
38.8%
Native American
19
2.5%
Pacific Islander
3
0.4%
Other
27
3.5%
Total
770
100%

SAMPLE 2
STOP
Frequency Percent
Gender
Male
30 25.2%
Female
89 74.8%
Total
119
100%
Age Groups
18-24 years
2
1.7%
25-34 years
18 15.1%
35-44 years
25 21.0%
45-54 years
38 31.9%
55-64 years
28 23.5%
65 years +
8
6.7%
Total
119
100%
Marital Status
Single
14 12.0%
Married
83 70.9%
Separated
1
0.9%
Divorced
17 14.5%
Widowed
2
1.7%
Total
117
100%
Race
White
106 90.6%
African American
2
1.7%
Hispanic
1
0.9%
Asian
3
2.6%
Native American
0
0.0%
Pacific Islander
0
0.0%
Other
5
4.3%
Total
117
100%

113

Table 3 (cont’d).
Family Structure
In a marriedcouple family
In a family with
female
householder, no
spouse present
In a family with
male householder,
no spouse present
In a group of
unrelated
subfamilies
Unrelated
individuals
Total
Recall experience
Have you
purchased a food
item that was later
recalled
Yes
No
Total
Have you ever
been diagnosed
with food borne
illness by a
medical
professional
Yes
No
Total

453

107

Family Structure
In a married59.2% couple family
In a family with
female
householder, no
14% spouse present

70

9.2%

30

3.9%

105
765

13.7%
100%

309
459
768

40.2%
59.8%
100%

183
583
766

23.9%
76.1%
100%

In a family with
male householder,
no spouse present
In a group of
unrelated
subfamilies
Unrelated
individuals
Total
Recall experience
Have you
purchased a food
item that was later
recalled
Yes
No
Total
Have you ever
been diagnosed
with food borne
illness by a
medical
professional
Yes
No
Total

114

86

74.8%

19

16.5%

6

5.2%

0

0.0%

4
115

3.5%
100%

60
58
118

50.8%
49.2%
100%

36
83
119

30.3%
69.7%
100%

Table 3 (cont’d).
Current
Occupation
Managerial or
professional
specialty
Executive,
administrative or
managerial
Professional
specialty
Technical, Sales, or
administrative
support
Technicians or
related support
Sales
Administrative
support or clerical
Service
Private household
Protective service
Service exc.
Protective, house
Farming, forestry,
or fishing
Precision
production, craft,
rep
Operations,
fabrication, or labor
Machine Operation,
assembly, or
inspect.
Transportation or
material moving
Handling or
cleaning
equipment, help,
lbr
Total

67

8.8%

79

10.4%

124

16.4%

44

5.8%

56
70

7.4%
9.2%

75
68
92
4

9.9%
9%
12.1%
0.5%

22

2.9%

9

1.2%

7

0.9%

11

1.5%

13

1.7%

10

1.3%

7
758

0.9%
100%

Current
Occupation
Managerial or
professional
specialty
Executive,
administrative or
managerial
Professional
specialty
Technical, Sales, or
administrative
support
Technicians or
related support
Sales
Administrative
support or clerical
Service
Private household
Protective service
Service exc.
Protective, house
Farming, forestry,
or fishing
Precision
production, craft,
rep
Operations,
fabrication, or labor
Machine Operation,
assembly, or
inspect.
Transportation or
material moving
Handling or
cleaning
equipment, help,
lbr
Total

115

21

18.8%

7

6.3%

36

32.1%

1

0.9%

4
3

3.6%
2.7%

10
4
18
2

8.9%
3.6%
16.1%
1.8%

1

0.9%

0

0.0%

3

2.7%

0

0.0%

0

0.0%

2

1.8%

0
112

0.0%
100%

Table 3 (cont’d).
Education
Less than High
School
High School/GED
Some College

5
58
147

Education
Less than High
0.7% School
7.6% High School/GED
19.1% Some College

0
5
21

0.0%
4.2%
17.8%

2 Year College
Degree

78

2 Year College
10.2% Degree

7

5.9%

4 Year College
Degree
Master’s Degree
Doctoral

296
162
12

4 Year College
38.5% Degree
21.1% Master’s Degree
1.6% Doctoral

46
30
2

39.0%
25.4%
1.7%

7
118

5.9%
100%

4
9
11
16
12
6
10
7
30
105

3.8%
8.6%
10.5%
15.2%
11.4%
5.7%
9.5%
6.7%
28.6%
100%

Professional (JD,
MD)
Total
Annual Income
Below $20,000
$20,000-$29,999
$30,000-$39,999
$40,000-$49,999
$50,000-$59,999
$60,000-$69,999
$70,000-$79,999
$80,000-$89,999
$90,000 or more
Total

10
768
237
157
101
94
65
37
32
14
32
769

Professional (JD,
1.3% MD)
100%
Total
Annual Income
30.9% Below $20,000
20.5% $20,000-$29,999
13.2% $30,000-$39,999
12.3% $40,000-$49,999
8.5% $50,000-$59,999
4.8% $60,000-$69,999
1.8% $70,000-$79,999
3.9% $80,000-$89,999
4.2% $90,000 or more
100%
Total

116

Table 4. Descriptive Statistics
Number

1
2

3

4

5
6
7
8
9
10

11

12

13
14

Item
The cantaloupe recall message was
believable.
The recall message emphasized
what might happen in the future if I
ate the cantaloupe.
The recall message emphasized
things that can happen immediately
if I ate the cantaloupe.
The recall message emphasized the
health benefits of not eating the
cantaloupe.
The recall message emphasized the
risks of eating the cantaloupe.
I am vulnerable to food borne
illness if I eat this cantaloupe.
I will get sick if I eat this
cantaloupe.
The likelihood of me getting sick
from eating this cantaloupe is.
I am at risk of getting sick if I ate
this cantaloupe.
Threats to my health if I bought and
ate this cantaloupe are
It could seriously injure me if I
didn't react to the instructions in this
recall message and ate the...
Protecting myself from injury based
on information in this cantaloupe
recall message is

It is within my control to protect
myself based on the information in
this cantaloupe recall message...
The primary responsibility for
protecting me from food borne
illness presented in this cantaloupe...

117

Mean S.D.

%
Disagree

% Agree

5.6 1.338

8.7%

84.1%

5.4 1.428

12.9%

81.4%

1.66

29%

55.7%

3.77 1.846

46.6%

38.6%

5.8 1.319

7.9%

87.6%

5.19 1.576

15.8%

74.4%

4.46

4.7 1.482

19.7%
61.2%
30.6%
57.2%
4.42 1.561 unlikely
likely
5.21 1.447

13.7%
77.7
19.5%
58.1%
below
above
4.72 1.534 average
average

4.97

1.47

17.7%

8.3%
5.53 1.353 unlikely

5.97 1.177

4.9%

5.65 1.361 9% myself

70.5%
81.9%
likely

89.6%
83%
myself

Table 4 (cont’d).
15

16

17

18

19

20
21

22
23
24

25
26

27

28

29

The primary responsibility for
protecting me from food borne
illness presented in this
cantaloupe...
Using the information in this
cantaloupe recall message will help
me prevent injury.
Responding to this cantaloupe
recall message is effective for
avoiding injury?
I am able to use the information in
this cantaloupe recall message to
prevent injury.
I would give up on responding to
this cantaloupe recall message
easily
Taking action based on this recall
notice is too complicated.
I will concentrate on ways to
reduce the risk of illness based on
this cantaloupe recall message.
I will make a plan of action to
protect myself when I read a recall
message like this one.
I will create solutions to avoid food
borne illness from recalled food…
I will think about the best way to
handle the risks from the….
I will concentrate my efforts on
doing what the cantaloupe recall
message suggests to avoid injury…
I will do what has to be done
according to this cantaloupe recall
message to avoid injury.
I will follow the information in this
cantaloupe recall message to avoid
injury.
I do not need to respond to this
cantaloupe recall message because I
am protected by my genes.
I believe that I am lucky enough to
avoid injury from this cantaloupe
recall.

118

46.7%
3.71 1.804 others

38.8%
others

5.61 1.255

6.5%

84.7%

5.53 1.301

8.5%

83.1%

5.6 1.331

8.3%

83.8%

3.57 1.884

50%

32.4%

5.6 1.331

66.3%

26.1%

5.41 1.308

8.4%

79.9%

5.59 1.301

7.5%

83.5%

5.34

1.33

9.3%

76.7%

5.51 1.243

7.7%

83.3%

5.64

1.28

6.7%

84.3%

5.8 1.165

4.9%

88%

5.84 1.198

6%

88.8%

2.44 1.637

76.3%

10.7%

3.39 1.977

53.2%

35.4%

Table 4 (cont’d).

30

31

32

33

34

35

36

37
38

39

40

41
42

43

I am unlikely to get sick if I eat the
cantaloupe in this recall message
because I have never had an...
I am better able to protect myself
and my family because of the
information presented in this
cantal...
Ignoring recall messages in the
future will increase my risk of
illness.
I am confident I will use recall
messages in the future to protect
my health.
I will use recall message
information in the future to reduce
the risk of serious illness from
eatin...
Using information in recall
messages is an effective means to
reduce the risk of serious illness.
This retailer has a great amount of
experience to create this cantaloupe
recall message.
This retailer has the skills to create
this cantaloupe recall message.
This retailer has enough experience
to create this cantaloupe recall
message.
This retailer does not have much
experience to create this cantaloupe
recall message. R
This retailer is making truthful
claims in this cantaloupe recall
message.
This cantaloupe recall message
from this retailer is honest.
I do not believe what this retailer is
telling me in this cantaloupe recall
message. R
I trust this cantaloupe recall
message from this retailer.

119

2.96

1.74

65%

22%

5.84 1.159

4.7%

89.2%

5.78 1.265

6.9%

77.5%

5.97 1.025

3.2%

87.5%

5.97 1.046

3.2%

92.5%

5.94 1.053

3.5%

91.1%

4.94 1.371

9.8%

59.6%

5.24 1.362

9.3%

72.7%

5.18 1.372

8.9%

70.7%

3.13 1.578

60.9%

17.5%

5.53 1.144

4.4%

80.7%

5.69 1.111

4.3%

85.5%

1.84

66.7%

22.6%

5.74 1.082

3.9%

88

2.98

Table 5A. Manipulation Check Mean Comparison
The recall
message
The
emphasized
Mean
cantaloupe what might
Condition
recall
happen in
message
the future if
was
I ate the
believable. cantaloupe.
Sample mean
5.60
5.40
Loss
5.64
5.52
condition
mean
Gain
5.63
5.42
condition
mean
Control
5.53
5.26
condition
mean

The recall
message
emphasized
things that
can happen
immediately
if I ate the
cantaloupe.
4.46
4.71

The recall
message
emphasized
the health
benefits of
not eating
the
cantaloupe.
3.77
3.98

The recall
message
emphasized
the risks of
eating the
cantaloupe.
5.80
5.89

4.45

3.89

5.83

4.22

3.44

5.68

Table 5B. Analysis of Variance Manipulation Check Combined Samples, Control, Gain,
and Loss Conditions
Item
The cantaloupe recall message was
believable.
The recall message emphasized what might
happen in the future if I ate the cantaloupe.
The recall message emphasized things that
can happen immediately if I ate the
cantaloupe.
The recall message emphasized the health
benefits of not eating the cantaloupe.
The recall message emphasized the risks of
eating the cantaloupe.

F

Sig.
.558
2.519

.081

6.584

.001

7.558

.001

2.006

120

.573

.135

Table 5C. Moderation Test of Predictor Variables Effects on Dependent Variables
Unconstrained Model
Variable(s)

Control

Severity----->Maladaptive Protection Motivations
Self-efficacy----->Maladaptive Protection Motivations
Locus of Control----->Maladaptive Protection
Motivations
Self-efficacy----->Adaptive Protection Motivations
Proactive Coping----->Adaptive Protection Motivations
Confidence in the Communicator-->Adaptive Protection
Motivations
Locus of Control----->Adaptive Protection Motivations
Reactive Coping----->Adaptive Protection Motivations

-.172**
.569***
.006
(n.s.)
.055*
.120*

Gain

Loss

.107(n.s.)

-.137*

-.615***

-.686***

-.174*
.087***
.116**

.063(n.s.)
.036(n.s.)
.202***

.284*** .192***
.129* .011(n.s.)
.197*** .326***

.250***
.158***
.219***

Full Constrained Model

Severity----->Maladaptive Protection Motivations
Self-efficacy----->Maladaptive Protection Motivations
Locus of Control (self)----->Maladaptive Protection
Motivations
Self-efficacy----->Adaptive Protection Motivations
Proactive Coping----->Adaptive Protection Motivations
Confidence in the Communicator-->Adaptive Protection
Motivations
Locus of Control----->Adaptive Protection Motivations
Reactive Coping----->Adaptive Protection Motivations
X2 Thresholds: *p<.05=53.61, **p<.01=54.99,
**p<.001=58.21.

Chisquare
Threshold

(n.s.)
(n.s.)
54.2

17

.233***
.066***
.259***

(n.s.)
55.1**
53.9*

17
17

ChiSquare
49
76
27

Variable(s)

Control
Gain
Loss
.141***
.625***
-.036
(n.s.)
.060***
.138***

df
15
31
16

df

(n.s.)
(n.s.)

X2 Difference Test
Overall Model
Unconstrained
Fully Constrained
Difference

121

p-value

.041*

Table 6. Model Fit Statistics, EFA and CFA
Model
EFA
CFA

2
Χ
df
p
SRMR RMSEA GFI
AGFI CFI
4607 666 <.001
0.08
0.078 0.77 0.73 0.812
1081 305 0.029 0.048
0.053 0.922 0.897 0.948

122

Table 7. Main Study Factor Loadings (EFA)

Number
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

Item
TASUS1
TASUS2
TASUS3
TASUS4
TASEV1
TASEV2
TASEV3
TASEV4
LOCUS1
LOCUS2
LOCUS
RESPON1
RESPON2
RESPON3
SELFEF1
SELFEF2
SELFEF3
COPING1
COPING2
COPING3
COPING4
COPING5
COPING6
MALPROT1
MALPROT2
MALPROT3
ADPROT1
ADPROT2
ADPROT3
ADPROT4
ADPROT5
CONFCOM1
CONFCOM2
CONFCOM3
CONFCOM4
CONFCOM5
CONFCOM6
CONFCOM7
CONFCOM8

Unstandardized
Loading
(Standard Standardized
2
R
Error)
Loading
.871 (.045)
0.641 0.411
.913 (.042)
0.714 0.509
.930 (.044)
0.691 0.478
1
0.803 0.644
.662 (.041)
0.552 0.305
.918 (.045)
0.674 0.455
.744 (.041)
0.612 0.374
1
0.766 0.587
1
0.789 0.623
.762 (.035)
0.687 0.472
.560 (.043)
0.439 0.193
.986 (.034)
0.851 0.724
.989 (.036)
0.825 0.681
1
0.815 0.664
2.455 (.213)
0.751 0.564
2.326 (.201)
0.741 0.549
1
0.412 0.17
.841 (.036)
0.67 0.449
.730 (.039)
0.57 0.325
.805 (.035)
0.673 0.454
1.027 (.032)
0.833 0.695
.959 (.028)
0.854 0.73
1
0.867 0.751
1
0.81 0.656
1.085 (.049)
0.728 0.53
1.038 (.043)
0.791 0.626
1
0.703 0.495
.966 (.054)
0.623 0.389
1.023 (.045)
0.813 0.66
1.060 (.046)
0.824 0.679
.981 (.045)
0.765 0.584
1
0.577 0.333
1.212 (.075)
0.705 0.496
1.200 (.075)
0.687 0.472
1.046 (.081)
0.51 0.26
1.067 (.064)
0.741 0.55
1.029 (.062)
0.738 0.545
.978 (.077)
0.502 0.252
1.054 (.061)
0.77 0.593
123

Table 8. Main Study Factor Loadings (CFA reduced model).

Number
7
8
9
11
12
13
14
15
17
18
19
20
21
23
24
25
26
27
28
29
30
31
34
35
36
41
42
44

Item
TASUS2
TASUS3
TASUS4
TASEV2
TASEV3
TASEV4
LOCUS1
LOCUS2
RESPON1
RESPON2
RESPON3
SELFEF1
SELFEF2
COPING1
COPING2
COPING3
COPING4
COPING5
COPING6
MALPROT1
MALPROT2
MALPROT3
ADPROT3
ADPROT4
ADPROT5
CONFCOM5
CONFCOM6
CONFCOM8

Unstandardized Loading (Standard Error)
1
1.043 (.052)
1.019 (.048)
.957 (.046)
.982 (.034)
1
1
.744 (.036)
.982 (.034)
.983 (.035)
1
1
.957 (.045)
1.074
1.047
1
.974 (.030)
.937 (.026)
1
1
1.088 (.049)
1.038 (.043)
1
1.035 (.036)
.941 (.037)
1
1.078 (.046)
1.086 (.045)

124

Standardized
Loading
0.736
0.73
0.77
0.703
0.628
0.767
0.788
0.67
0.85
0.823
0.818
0.769
0.77
.763
.728
.763
0.817
0.862
0.895
0.81
0.729
0.791
0.831
0.841
0.767
0.741
0.824
0.846

2
R
0.542
0.533
0.594
0.494
0.394
0.588
0.621
0.448
0.723
0.677
0.668
0.591
0.593
.582
.530
.556
0.667
0.743
0.801
0.656
0.531
0.625
0.69
0.707
0.588
0.549
0.679
0.716

Table 9. Item and Scale Reliabilities, Reduced Model (CFA)

Number
7
8
9
10
11
12
14
15
17
18
19
20
21
23
24

Item Loading AVE Scale
TASUS2
0.736 0.541
TASUS3
0.731 0.535
TASUS4
0.770 0.592 Susceptibility
TASEV2
0.767 0.493
TASEV3
0.788 0.395
TASEV4
0.67 0.588 Severity
LOCUS1
0.628 0.621
LOCUS2
0.767 0.448 Locus of Control
RESPON1
0.791 0.723
RESPON2
0.831 0.678
RESPON3
0.841 0.668 Response Efficacy
SELFEF1
0.81 0.591
SELFEF2
0.729 0.594 Self-efficacy
COPING1
0.763 0.582
COPING2
0.728 0.530

25
26
27
28
29
30

COPING3
COPING4
COPING5
COPING6
MALPROT1
MALPROT2

0.746
0.829
0.865
0.884
0.769
0.77

31
34
35
36
41
42
44

MALPROT3
ADPROT3
ADPROT4
ADPROT5
CONFCOM5
CONFCOM6
CONFCOM8

0.85
0.823
0.818
0.703
0.736
0.73
0.77

0.556 Pro-active Coping
0.687
0.748
0.782 Reactive Coping
0.655
0.531
Maladaptive Protection
0.625 Motivation
0.697
0.706
0.583 Adaptive Protection Motivation
0.549
0.678
0.716 Confidence in Communicator

125

Cronbach's
Alpha

0.790

0.745
0.69

0.868
0.744

0.791

.893

0.821

0.852

0.846

Table 10A. Manipulation Check Structural Formative Model for
Theorized Protection Motivation Dimensions (Control Condition)
CONTROL

Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check

Estimate

<--- Susceptibility
<--- Severity
Response
<--- Efficacy

<--- Self-efficacy
Pro-active
<--- Coping
Reactive
<--- Coping
Confidence in
Manipulation
the
Check
<--- Communicator
Manipulation
Locus of
Check
<--- Control
***p<.001, **p<.01, *p<..05

S.E.

C.R.

P

.216 0.057 3.764

***

0.073 0.071 1.029 0.304
0.111 0.072 1.534 0.125
-0.132 0.039

-3.42

***

.069

-.898

.369

-0.077 0.083

-.932

.351

0.244 0.066

3.70

***

0.138 0.078

1.78 0.076

-.062

126

Table 10B. Manipulation Check Structural Formative Model for
Theorized Protection Motivation Dimensions (Gain Condition)
GAIN
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check

Estimate

S.E.

C.R.

P

<--- Susceptibility

0.089 0.058 1.538 0.124

<--- Severity
Response
<--- Efficacy

0.075 0.063 1.190 0.234

<--- Self-efficacy
Proactive
<--- Coping
Reactive
<--- Coping
Confidence in
Manipulation
the
Check
<--- Communicator
Manipulation
Locus of
Check
<--- Control
***p<.001, **p<.01, *p<..05

0.171 0.065 2.635

**

0.065 0.040 1.646 0.100
0.071 0.064 1.099 0.272
0.035 0.081

.435 0.663

0.118 0.068 1.732 0.083
0.006 0.071 0.092 0.927

127

Table 10 C. Manipulation Check Structural Formative Model for
Theorized Protection Motivation Dimensions (Loss Condition)
LOSS
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check
Manipulation
Check

Estimate

S.E.

C.R.

P

<--- Susceptibility

0.030 0.057 0.533 0.594

<--- Severity
Response
<--- Efficacy

0.133 0.062 2.146

<--- Self-efficacy
Pro-active
<--- Coping
Reactive
<--- Coping
Confidence in
Manipulation
the
Check
<--- Communicator
Manipulation
Locus of
Check
<--- Control
***p<.001, **p<.01, *p<..05

-0.005 0.061
-.092
.174

.035

-.082 0.935
-2.62

**

.061 2.871

**

.131 0.075

-1.75 0.081

0.096 0.062 1.541
0.061 0.069

128

*

.123

.884 0.377

Table 11A. Significant Multivariate Effects, Multivariate Analysis of Variance (MANOVA)
Wilk's
Lambda
Variable(s)
Susceptibility
Severity
Response Efficacy
Self- Efficacy
Proactive Coping
Reactive Coping
Locus of Control
Confidence in the Message
Communicator
Treatment (control, gain, loss)

1
0.988
0.994
0.643
0.968
0.925
0.989

F
0.18
5.393
2.53
252.0
15.24
36.89
5.24

2
2
2
2
2
2
2

0.907 46.80
0.997 0.67

2
4

Table 11B. Box's Test of Equality of Covariance
Matrices
Box's M
F
df (1)
df (2)
Significance

3.751
0.623
6
2084
0.712

Table 11C. Levene's Test of Equality of Error
Variances
Variable(s)
Maladaptive Protection Motivations
Adaptive Protection Motivations

df

df
F
(1)
0.655
2
0.162
2

129

df
(2)
918
918

p
0.52
0.85

Error
df
909
909
909
909
909
909
909

p
n.s.
0.005**
n.s.
.000***
.000***
.000***
.005**

909 .000***
1818
n.s.

Table 12. Hypothesis Test for Structural Model Path Analysis
Dependent
Variable
Maladaptive
Protection
Motivation
Maladaptive
Protection
Motivation
Maladaptive
Protection
Motivation
Adaptive
Protection
Motivation

Independent
Variable

β

<--

Control
Susceptibility

.072

.073

.983

<-<--

Control
Severity
Control/Gain/
Loss
Self-efficacy

-.226
-.579
-.621
-.688

.091
.049
.048
.045

-2.48
-11.7
-12.92
-15.39

*** 3 Supported

<--

Gain
Self-efficacy

.091

.027

3.424

*** 4 Supported

Control/Gain/
Loss
Response
Efficacy

.019
-.069
.012

.060
.051
.049

.324
-1.357
.241

Maladaptive
Protection
<Motivation --

S.E.

130

CR

P

HYPOTHESIS

n.s. 1 Not Supported

* 2 Supported

n.s. 5 Not Supported

Table 12 (cont’d).

Adaptive
Protection
Motivation
Adaptive
Protection
Motivation
Adaptive
Protection
Motivation

Adaptive
Protection
Motivation

Adaptive
Protection
Motivation
Adaptive
Protection
Motivation
Adaptive
Protection
Motivation

Adaptive
Protection
Motivation

<--

<--

<--

Control/Gain/
Loss
Response
Efficacy
Control/Gain/
Loss
Pro-active
Coping
Control/Gain/
Loss
Reactive
Coping
Moderation
Test (Tbl. 5B)

<--

Loss-->
Pro-active
Coping
Moderation
Test (Tbl. 5B)
Gain -->
Reactive
Coping
Loss
Locus of
Control
Moderation
Test (Tbl. 5B)
Loss Locus
of Control

<--

Control/Gain/
Loss
Confidence in
the Message
Communicator

<--

<-<--

.059
-.065
-.033

.049
.043
.042

1.203
-1.531
-.782

.127
.095
.195

.047
.043
.042

2.683
2.233
4.652

** 7 Supported

.183
.357
.223

.056
.053
.052

3.247
6.672
4.334

*** 8 Supported

n.s. 6 Not Supported

9 Supported

10 Supported

.144

.048

3.009

** 11 Supported

12 Supported

.277
.195
.250

.045
.045
.043

131

6.114
4.284
5.818

*** 13 Supported

2
Table 13. R Summary by Experimental Condition
2
2
R
R
Adaptive
Maladaptive
Control
0.539
0.415
Gain
0.557
0.471
Loss
0.605
0.511

132

Table 14. MTurk Participant Experiment Comments Post-test
POSITIVE /QUALITATIVE
COMMENTS
It was a good survey to be part of. I
am generally very likely to catch
disease because of my low immunity.
So I don’t take any kind of risks
related to food products as I had a
previous experience of getting sick
due to food products.

Completed cantaloupe recall message
survey. It was a very well thought out
survey.
We take such recalls seriously. We
would return or throw out possibly
contaminated food.
I was affected by the Listeria
outbreak last summer. The
cantaloupes in question were from
Colorado. Nasty stuff, Listeria. Made
me hesitant to eat cantaloupes ever
since.
For whatever cynical perspective I
have, I would think the producers
would attempt to disguise or hide
from food contamination charges if
they could, and only bring them to
light as a consequence of exposure
via media, etc.
illness from food borne pathogens,
now that just freaks me out. I use to
pay really close attention to recalls
when I had a new born, cause baby
products seem to get recalled a lot.
Food recalls are not as common and
the totally freak me out.

NEGATIVE COMMENTS
Most boring, tedious, pointless study
ever! Talk about over-complicating
things! Is a simple recall notice so
unfathomable to the average American
that you need to ask the same questions
so many slightly re-worded ways to
make sense of it? Jee-zus.....
I found the use of the term 'injury' to be
confusing. After several questions with
injury in it, I realized that you must be
using it as a synonym for illness. But
the first few questions with that term I
answered false, as I deem injury to be
more direct physical damage. You
don't get injured from the flu, for
example, although you do suffer an
illness.
this survey was very long. It asked the
same questions over and over. I think u
would get more honest answers if u cut
out half the questions in this survey

The questions got a little repetitive and
it seemed like you were asking for the
same response in several questions.

survey seemed very wordy!! Asking the
same question in 10 different ways.

Questions were very repetitive - if it's
not necessary to ask the same thing 4
different ways for your research, I
would strongly suggest cutting it down.

133

Table 14 (cont’d).
due to the nature of the threat of the
recall identifying pregnant women as
more vulnerable, it might have been a
good survey question to ask if the
survey taker is pregnant or has young
children
This is A good survey. So far I have
taken many surveys and this one of
the best and an interesting one. Its
very important to buy food items
which are safe and not contaminated.
This was a well-presented survey,
easy to understand. The questions
were appropriate and there was
sufficient variety in the responses
provided. It was not at all confusing,
and I like the survey.
Great survey, just the exact amount of
questions per page to make the
survey easy to understand and do.
I buy only organic fruits and
vegetables. I ALWAYS wash my
fruits/veggies first before we eat
them. I use extra precautions like this
for this exact purpose. Better safe
than sorry:)
I worked in a restaurant for about 6
months. Some of the food handling
procedures were useful. Others were
stupid. Like pouring bleach on
leftover food. This was supposed to
keeps bums from taking it out of the
dumpster.

Interesting survey--hard to believe
people would ignore such warnings.

Jesus Christ guys, you ask the same god
damn questions 15-20 different ways
each
Many of the questions were very
repetitive. They were just worded
slightly differently from each other.
Sometimes the same question was
asked 4 times in a row.

probably entered this like it says in the
directions, but there were so many
duplicate questions I don't remember
now.
Perhaps the most annoying study I have
ever taken, due to its format.
I had problems with the recall notice. I
do not feel like it addressed how the
recall should be addressed with the
normal individual. Only if you were
pregnant or had immune problems
I thought that a number of the sentences
kept repeating themselves-they were
phrased differently but said essentially
the same thing. Also, the English in
some of the sentences sounded a little
awkward and not quite natural.
Otherwise, though, I thought the survey
was fine.
Some of the questions were worded a
little strangely. I had to read them
several times to make sure I understood
what was being asked.

134

APPENDIX B
FIGURES
Figure 1. Conceptual Diagram of Prospect Theory and Protection Motivation Theory
Experiment

135

Figure 2A. Informed Consent Message
Research Participant Information and Consent Form
1. EXPLANATION OF THE RESEARCH AND WHAT YOU WILL DO:



2.

You are being asked to participate in a research study of your perceptions food product
recall messages
The research study were conducted (online). You were asked to read a recall message
followed by a short survey.
You must be 18 years of age to participate in this research study.
YOUR RIGHT TO PARTICIPATE, SAY NO, OR WITHDRAW:



Participation in this research project is completely voluntary. You have the right to say
no. You may change your mind at any time and withdraw. You may choose not to
answer specific questions or to stop participating at any time. Whether you choose to
participate or not will have no effect on your grade or evaluation.

3. COMPENSATION FOR BEING IN THE STUDY:
 A random drawing of five $100 VISA gift cards were conducted at the end of the study
and you may win. You were notified by email if you are a winner and a request for your
shipping information to receive the gift card. You must complete the survey to be entered
into the gift card drawing.
4. CONTACT INFORMATION FOR QUESTIONS AND CONCERNS

If you have concerns or questions about this study such as scientific issues, how to do any part of
it, or to report an injury, please contact the researcher (Greg Clare, 314 Communication Arts and
Sciences, East Lansing, MI 48824, claregre@msu.edu, 517-353-3299. You may also contact the
principal investigator Patricia Huddleston @ huddles2@msu.edu.
By clicking “Yes” below you acknowledge that you have read and understand that:


Your participation in this survey is voluntary. You may withdraw your consent
and discontinue participation in the project at any time. Your refusal to participate
will not result in any penalty.



You have given voluntary agreement to participate in this research.

Do you wish to participate in this study?
__ Yes, I want to participate

__ No, I do not want to participate
136

Figure 2B. Invitation and Reminder Message

Research Study - Your chance to win $100 in one of five drawings
Dear _______________
I am a PhD candidate at Michigan State University in the Advertising, Public Relations and
Retailing Department. Below is a link to an online experiment that is part of my doctoral
research. The experiment will take approximately 15-20 minutes and is about food product
recalls. By completing the experiment you were entered into a drawing. Five winners were
randomly selected to receive a Visa gift card of $100. Your participation is greatly appreciated!
URL
Thank you for your time,
Greg Clare
This link is unique to you. Please do not forward it.
Reminder Message:
YOUR CHANCE TO WIN $100 IN THE NEXT FEW WEEKS.
Dear [Name],
I am conducting an online experiment for my doctoral research at Michigan State University in
the Advertising, Public Relations and Advertising Department. I would appreciate your help. I
sent you an email invitation on DATE, but haven't heard from you yet. I sincerely value your
insights into food recall messages and preventing the risks of food borne illness for people just
like you in your community.
The survey should take no more than 20 minutes to complete, and you may just win a $100 Visa
gift card for your time. To access our experiment, please click the link provided below.
To unsubscribe to our email list and future surveys, please send an email to claregre@msu.edu
Thank you for your time and we look forward to hearing from you!
Best,
Greg Clare
This link is unique to you. Please do not forward it.

137

Figure 3A. FDA Recall Press Release of Stimulus Cantaloupe Recall Message
FDA PRESS RELEASE
For Immediate Release: Sept. 14, 2011
Media Inquiries: Siobhan DeLancey, 301-796-4668, siobhan.delancey@fda.hhs.gov
Consumer Inquiries: 888-INFO-FDA
FDA warns consumers not to eat Rocky Ford Cantaloupes shipped by Jensen Farms
Jensen Farms recalls Rocky Ford cantaloupe due to potential link to a multi-state outbreak of
listeriosis
Fast Facts








The FDA is warning consumers not to eat Rocky Ford Cantaloupe shipped by Jensen
Farms and to throw away recalled product that may still be in their home.
Jensen Farms is voluntarily recalling Rocky Ford Cantaloupe shipped from July 29
through September 10, 2011, and distributed to at least 17 states with possible further
distribution.
The recalled cantaloupes have the potential to be contaminated with Listeria and may be
linked to a multi-state outbreak of listeriosis.
The CDC reports that at least 22 people in seven states have been infected with the
outbreak-associated strains of Listeria monocytogenes as of September 14.
Patients reported eating whole cantaloupes they purchased from grocery stores marketed
from the Rocky Ford growing region of Colorado.
While all people are susceptible to Listeria, older adults, persons with weakened immune
systems and pregnant women are at particular risk.

What is the Problem?
The FDA is warning consumers not to eat Rocky Ford Cantaloupe shipped by Jensen Farms of
Granada, Colo. The majority of the patients reported eating cantaloupe marketed from the Rocky
Ford growing region. FDA’s trace back data from the State of Colorado about their confirmed
cases of Listeria monocytogenes have identified a common producer of Rocky Ford cantaloupes.
That producer is Jensen Farms. Although the investigation is ongoing, no other Rocky Ford
cantaloupe producer has been found in common in the Colorado trace back.
Jensen Farms is voluntarily recalling Rocky Ford Cantaloupe. The recalled cantaloupes were
shipped from the Rocky Ford growing region of Colorado from July 29 through September 10
and are potentially linked to a multi-state outbreak of listeriosis. The recalled cantaloupes were
distributed to at least 17 states with possible further distribution.

138

Figure 3A (cont’d).
What are the Symptoms of Listeriosis?
Listeriosis is a rare and serious illness caused by eating food contaminated with bacteria called
Listeria. Persons who think they might have become ill should consult their doctor. A person
with listeriosis usually has fever and muscle aches.
Who is at Risk?
Listeriosis can be fatal, especially in certain high-risk groups. These groups include older adults,
people with compromised immune systems and certain chronic medical conditions (such as
cancer), and unborn babies and newborns. In pregnant women, listeriosis can cause miscarriage,
stillbirth, and serious illness or death in newborn babies, though the mother herself rarely
becomes seriously ill.
What Do Consumers Need To Do?
Consumers should not eat Rocky Ford Cantaloupe shipped by Jensen Farms and should
immediately discard the recalled cantaloupes in the trash in a sealed container so that children
and animals, such as wildlife, cannot access them. Consumers who are concerned about illness
from Listeria monocytogenes should consult their healthcare professionals.
What Does the Product Look Like?
The cantaloupe may be labeled: Colorado Grown, Distributed by Frontera Produce, USA,
Pesticide Free, Jensenfarms.com, Sweet Rocky Fords.
http://www.fda.gov/Safety/Recalls/ucm271882.htm
The cantaloupes are packed in cartons that are labeled: Frontera Produce,
www.fronteraproduce.com or with Frontera Produce, Rocky Ford Cantaloupes. Both cartons also
include: Grown and packed by Jensen Farms Granada, CO and Shipped by Frontera Produce
LTD, Edinburg, Texas.
Not all of the recalled cantaloupes are labeled with a sticker. Consumers should consult the
retailer if they have questions about the origin of a cantaloupe.
Where is it Distributed?
The recalled cantaloupes were distributed to the following states: IL, WY, TN, UT, TX, CO,
MN, KS, NM, NC, MO, NE, OK, AZ, NJ, NY, PA. Further distribution is possible.
What is Being Done about the Problem?
Jensen Farms is working with the FDA and the State of Colorado to remove its Rocky Ford
Cantaloupe from the marketplace. The FDA is also working with CDC, the states and other
regulatory partners to investigate where in the supply chain the contamination occurred.
This is the first time a Listeria monocytogenes outbreak has been reportedly linked to whole
cantaloupe. Foods that typically have been associated with food borne outbreaks of Listeriosis
are deli meats, hot dogs, and Mexican-style soft cheeses made with unpasteurized milk.
Listeriosis has not often been associated with the consumption of fresh produce with the
139

Figure 3A (cont’d).
exception of two food borne illness outbreaks related to consumption of sprouts in 2009 and
fresh-cut celery in 2010.
Because of this unusual circumstance, FDA’s newly formed Coordinated Outbreak Response and
Evaluation (CORE) Network is working with FDA Districts, CDC, the States and other
regulatory partners on a root cause analysis to determine where in the supply chain and what
circumstances likely caused the implicated cantaloupe to be contaminated. FDA is exploring
whether harvesting and/or postharvest practices may have contributed to this contamination, as
well as what could be done differently to prevent future occurrences.

For more information:
CDC Investigation on multi-state listeriosis outbreak:
http://www.cdc.gov/nczved/divisions/dfbmd/diseases/listeriosis/outbreak.html
Listeria page on FS.gov:
http://www.foodsafety.gov/poisoning/causes/bacteriaviruses/listeria/index.html
Produce Safety page on FDA:
http://www.fda.gov/Food/ResourcesForYou/Consumers/ucm114299
Coordinated Outbreak Response and Evaluation (CORE) Network:
http://www.fda.gov/Food/FoodSafety/CORENetwork/default.htm

140

Figure 3B. Control Group Message Stimulus
SWEET ROCKY FORD CANTALOUPE - KROGER FARMS , VARIOUS SIZES
Affected in Kroger stores.
Reason: The cantaloupe may be contaminated with Listeria monocytogenes and, if eaten, could
result in severe illness to those individual who are pregnant or have a weakened immune system.

Figure 3C. Gain Group Message Stimulus
SWEET ROCKY FORD CANTALOUPE - KROGER FARMS , VARIOUS SIZES
Affected in Kroger stores.
Reason: The cantaloupe may be contaminated with Listeria monocytogenes and, if eaten, could
result in severe illness, especially to those individuals who are pregnant or have a weakened
immune system.
80% of people infected with Listeria monocytogenes will not become seriously ill, but to ensure
your safety please throw out any cantaloupe you purchased at Kroger immediately.
February 1, 2012

Figure 3D. Loss Message Group Stimulus
SWEET ROCKY FORD CANTALOUPE - KROGER FARMS , VARIOUS SIZES
Affected in Kroger stores.
Reason: The cantaloupe may be contaminated with Listeria monocytogenes and, if eaten, could
result in severe illness to those individual who are pregnant or have a weakened immune system.
20% of people infected with Listeria monocytogenes will become seriously ill, but to ensure
your safety please throw out cantaloupe immediately.
February 1, 2012

141

Figure 4. Graphic Stimuli (Control, Gain, Loss) Conditions

For interpretation of the references to color in this and all other figures, the reader is referred to
the electronic version of this dissertation.

142

Figure 5. Main Study CFA Model (reduced)

143

Figure 6A. Manipulation Check Control Condition

(B)*** p<.001, **p<.01 **, p<.05 *
144

Figure 6B. Manipulation Check Gain Condition

(B)*** p<.001, **p<.01 **, p<.05 *

145

Figure 6C. Manipulation Check Loss Condition

(B)*** p<.001, **p<.01 **, p<.05 *

146

Figure 7A. Structural Model Path Diagram Control Condition

(B)*** p<.001, **p<.01 **, p<.05 *

147

Figure 7B. Structural Model Path Diagram Gain Condition

(B)*** p<.001, **p<.01 **, p<.05 *
148

Figure 7C. Structural Model Path Diagram Loss Condition

(B)*** p<.001, **p<.01 **, p<.05 *
149

BIBLIOGRAPHY

150

BIBLIOGRAPHY

Aaker, D. A. & Keller, K.L. (1990). Consumer evaluations of brand extensions, Journal of
Marketing, 54, 27-41.
Anderson, J.C. & Gerbing, D.W. (1988). Structural equation modeling in practice: A review
and recommended two-step approach, Psychological Bulletin, 103(3), 411-423.
Arora, R. (2007). Message framing strategies for new and mature products, Journal of Product
& Brand Management, 16(6), 377-385.
Arora, R., Stoner, C., & Arora, A. (2006). Using framing and credibility to incorporate exercise
and fitness in individuals’ lifestyle, Journal of Consumer Marketing, 23(4), 199-207.
Bandura, A. (1977). Self-efficacy: Toward a Unifying Theory of Behavior Change,
Psychological Review, 84(2), 191-215.
Bandura, A. (1999). Social cognitive theory: An agentic perspective, Asian Journal of Social
Psychology Special Issue: Theoretical and methodological advances in social
psychology, 2(1), 21-41.
Bandura, A. (2000). Self-efficacy: The foundation of agency, In Perrig, W.J., Alexander, G.
(eds.) Control of human behavior, mental processes, and consciousness. New Jersey:
Erlbaum, 17-33.
Banks, S. M., Salovey, P., Greener, S., Rothman, A., Moyer, A., Beauvais, J. & Epel, E. (1995).
The effects of message framing on mammography utilization, Health Psychology, 14,
178-184.
Beck, K. & Lund, A. (1981). The effects of health threat seriousness and personal efficacy upon
intentions and behavior, Journal of Applied Social Behavior, 11(5), 401-415.
Bender, H., Martin, I.M. & Raish, C. (2006). What motivates homeowners to protect themselves
from wildfire risks in the WUI, In Wildfire and fuels management: Risk and human
reaction, Ingrid Martin, Ed. Long Beach: California State University Press, 1-34.
Block, L. G. & Keller, P.A. (1995). When to accentuate the negative: The effects of perceived
efficacy and message framing on intentions to perform a health related behavior, Journal
of Marketing Research, 32, 192-203.
Boer, H., & Seydel, E. R. (1996). Protection motivation theory. In M. Conner & P. Norman
(Eds.), Predicting health behavior: Research and practice with social cognition models,
(pp. 95-120). Buckingham: Open University Press.

151

Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation
perspective, Psychological Bulletin, 110, 3005-3014.
Consumer Product Safety Commission, (1978). Recall effectiveness study, retrieved from
http://www.cpsc.gov/index.html
Consumer Product Safety Commission, (1980). Report of the recall effectiveness task force of
the Consumer Product Safety Commission, retrieved from:
http://www.cpsc.gov/index.html
Consumer Product Safety Commission, (2011). FAQ page, retrieved from
http://www.cpsc.gov/index.html
Daly, M.C., Duncan, G.J, McDonough, P., & Williams, D. (2000). Optimal indicators of
socioeconomic status for health research, Institute for Policy Research, Northwestern
University Press, Evanston, 1-27.
Dawson, C. R. (2005). Food scares and food safety regulation: Qualitative research on current
public perceptions, Food Standards Agency, London.
Detweiler, J.B., Bedell, B.T., Salovey, P., Pronin, E., & Rothman, A.J. (1999). Message framing
and sunscreen use: Gain-framed messages motivate beach-goers, Health Psychology, 18,
189-196.
Donovan, R. J. & Jalleh, G. (2000). Positive versus negative framing of a hypothetical infant
immunization: The influence of involvement, Health Education & Behavior, 27(1), 8295.
Donovan, R. J. & Jalleh, G. (1999). Positively versus negatively framed product attributes: The
influence of involvement, Psychology & Marketing, 16(7) 613-630.
Dorn, L. & Brown, B. (2003). Making sense of invulnerability at work: A quantitative study of
police drivers, Safety Science, 41, 837-859.
Duhachek, A. (2005). Coping: A multidimensional, hierarchical framework of responses to
stressful consumption episodes, Journal of Consumer Research, 32, 41-53.
Eagley, A. H. & Chaiken, S. (1984). Cognitive theories of persuasion, Advances in
Experimental Social Psychology, 17, 267-359.
Farell, A.M., & Rudd, J.M. (2009). Factor analysis and discriminant validity: Some practical
issues, proceedings of the Australian and New Zealand Social Marketing Conference,
retrieved from http://www.duplication.net.au/ANZMAC09/papers/ANZMAC2009389.pdf

152

Floyd, D.L, Prentice-Dunn, S., & Rogers, R.W. (2000). A meta-analysis of research on
protection motivation theory, Journal of Applied Social Psychology 30 (2), 407-429.
Food Poisoning Law Blog (2011). Prizker Olson Attorneys. Available from:
http://foodpoisoning.pritzkerlaw.com/archives/cat-michigan-e-coli.html
Food Processing. 2007. Survey reveals eroding consumer confidence in packaged goods brands.
Putman Media. Available from:
http://www.foodprocessing.com/industrynews/2007/093.html. Accessed Dec 20, 2008.
Food Product Design (2008). Survey illustrates consumer food safety fears,
http://www.foodproductdesign.com/hotnews/survey-illustrates-consumer-food-safetyfears.html
Fornell, C., & Larcker, D. (1981). Evaluating structural equation models with unobservable
variables and measurement error, Journal of Marketing Research, 18 (3), 39-50.
Fox, M., and P. Simao. 2009. Many Americans ignore food recalls: survey. Reuters, Tue Apr 14.
http://www.reuters.com/article/domesticNews/idUSTRE53D33E20090414. Accessed
April 14, 2009.
FSIS. (2009).United States Department of Agriculture, Food Safety Inspection Service. Recall
Case Archive.
(http://www.fsis.usda.gov/Fsis_Recalls/Recall_Case_Archive/index.asp).Accessed April
13, 2009.
Gefen, D., & Straub, D. (2005). A practical guide to factorial validity using PLS-Graph:
Tutorial and annotated example, Communications of the Association for Information
Systems, 16, 91-109.

Goldberg, M.E., Hartwick, J. (1990). The effects of advertiser reputation and extremity of
advertising claim on advertising effectiveness, Journal of Consumer Research, 17, 172179.
Haberkorn, J. 2007. Food recalls damage consumers’ confidence in safety. Business News, The
Washington Times. Available from: http://washingtontimes.com/news/2007/jul/05/foodrecalls-damage-consumers-confidence-in-safety/. Accessed Sept 10, 2009.
Hallman, W.K., Cuite, C.L. & Hooker, H.H. (2009, 14 April). Consumer Responses to Food
Recalls: 2008 National Survey Report. (Publication Number RR-0109-018). New
Brunswick, New Jersey: Rutgers, the State University of New Jersey, Food Policy
Institute.
Harrington, S.J. (1996). The effect of codes of ethics and personal denial of responsibility on
computer abuse judgments and intentions, MIS Quarterly, 20, 257-258.

153

Hemphill, T. (2009). Globalization of the U.S. Food Supply: Reconciling Product Safety
Regulation with Free Trade. Business Economics, 44(3), 154-184.
Ho, R., Davidson, G., & Ghea, V. (2005). Motives for the adoption of protective health
behaviors for men and women: An evaluation of the psychosocial-appraisal health
model, Journal of Health Psychology, 10, 373-395.
Jarvis, C.B., Mackenzie, S.B., & Podsakoff, P.M. (2003). A critical review of construct
indicators and measurement model misspecification in marketing and consumer research,
Journal of Consumer Research, 30, 199-218.
Jones, L.W., Sinclair, R.C., & Courneya, K.S. (2003). The effects of source credibility and
message framing on exercise intentions, behaviors, and attitudes: An integration of the
elaboration likelihood model and prospect theory, Journal of Applied Social Psychology,
33(1), 179-196.
Kahneman, D. & Tversky, A. (1979). Prospect theory: Analysis of decision under risk,
Econometrica, 47(2), 263-291.
Kalichman, S.C., & Coley, B. (1995). Context framing to enhance HIV-antibody-testing
messages targeted to African American women, Health Psychology, 14, 247-254.
Keller, P. A., & Block, L. G. (1998). Beyond protection motivation: An integrative theory of
health appeals, Journal of Applied Psychology, 28, 1584-1608.
Keller, P. A., & Block, L. G. (1996). Increasing the persuasiveness of fear appeals: The effect of
arousal and elaboration. Journal of Consumer Research, 22, 448–459.
Keller, P.A., & Block, L.G. (1995). When to accentuate the negative: The effects of perceived
efficacy and message framing on intentions to perform a health-related behavior, Journal
of Marketing Research, 32(2), 192-203.
Kline, R. B. (2004). Principles and practice of structural equation modeling, 2nd Ed. New York:
Guilford.
Knobloch-Westerwick, S; Carpentier, FD; Blumhoff, A; Nickel, N. (2005). Selective exposure
effects for positive and negative news: Testing the robustness of the informational utility
model. Journalism & Mass Communication Quarterly, 82 (1): 181-195 SPR 2005
Krantz, J.H. & Dalal, R. (2000). Validity of web-based psychological research. In M. H.
Birnbaum (Ed.), Psychological experiments on the Internet, 35-60. San Diego:
Academic Press.
Kroger Stores, (2012). Kroger Services Recall Information, retrieved from
www.kroger.com/services/pages/recall_information.aspx.

154

LaBarbera, P.A. (1982). Overcoming a no-reputation liability through documentation and
advertising documentation, Journal of Marketing Research, 19, 223-228.
LaRose, R., Rifon, N.J., & Enbody, R. (2008). Promoting personal responsibility for Internet
safety, Communications of the ACM, 51(3), 71-76.
Laufler D. & Coombs, W. (2006). How should a company respond to a product harm crisis?
The role of corporate reputation and consumer-based cues. Business Horizons 49, 379385.
Lazarus, R.S., & Folkman, S.E. (1984). Stress, Appraisal and Coping, Springer: New York,
NY.
Levene, H. (1960). Robust tests for equality of variance, In I. Olkin, S. Ghurye, W. Hoeffding,
W. Maddow, & H. Mann Eds., Contributions to probability and statistics: Essays in
honor of Howard Hotelling, Stanford University Press: Stanford, CA, 278-291.
Lynch, J. G. (1982). On the external validity of experiments in consumer research, Journal of
Consumer Research, 9, 225-239.
Mackenzie, SB & Lutz, R.J. (1985). An empirical examination of the structural antecedents of
attitude toward the ad in an advertising pretesting context, Journal of Marketing, 53, 4865
Maddux, J. E. & Rogers, R. (1982). Protection motivation and self-efficacy: A revised theory of
fear appeals and attitude change, Journal of Experimental Social Psychology, 19(5), 469479.
Maheswaran, D., & Meyers-Levy, J.M. (1990). The influence of message framing and issue
involvement, Journal of Marketing Research, 27, 361-367.
Marteau, T.M. (1989). Framing of information: Its influence upon decisions of doctors and
patients, Journal of Social Psychology, 28, 89-94.
Martin, D. W. (1996). Doing psychology experiments. (4th ed.). Pacific Grove, CA:
Brooks/Cole.
Matos, C. & Rossi, C. (2007). Consumer reaction to product recalls: factors influencing product
judgment and behavioral intentions, International Journal of Consumer Studies, 31, 109116.
McNeil, B. J., Pauker, S. G., Sox, H.C., & Tversky, A. (1982). On the elicitation of preferences
for alternative therapies, New England Journal of Medicine, 306, 1259-1262.

155

Mead, P., Slutsker, L., Dietz, V., McCaig, L., Bressee, J., Shapiro, C., Griffin, P. & Tauxe, R.
(1999). Food-related illness and death in the United States, Emerging Infectious Diseases,
5, 607-625.
Menon, G., Block, L. & Ramanathan, S. (2002). We're at as much risk as we are led to believe:
Effects of message cues on judgments of health risk, Journal of Consumer Research, 28
(4), 533-549.
Meyerowitz, B.E. & Chaiken, S. (1987). The effect of message framing on breast selfexamination attitudes, intentions, and behavior, Journal of Personality and Social
Psychology, 52(3), 500-510.
Milne, S., Sheeran, P. and Orbell, S. (2000). Prediction and Intervention in Health-Related
Behavior: A Meta-Analytic Review of Protection Motivation Theory, Journal of Applied
Psychology, 30(1) 106-143.
Milne, S., Sheeran, P. and Orbell, S. (2002). Combining motivational and volitional
interventions to promote exercise participation: Protection motivation theory and
implementation intentions, British Journal of Health Psychology, 7, 163-184.
Myers, R.E. (2010). Moderating the effectiveness of messages to promote physical activity in
type 2 diabetes (Doctoral dissertation), Retrieved from University of South Florida
Scholar Commons, Theses and Dissertations. Paper 1719.
http://scholarcommons.usf.edu/etd/1719
Neuwirth, K., Dunwoody, S. & Griffin, R.J. (2000). Protection motivation and risk
communication, Risk Analysis, 20 (5), 721-734.
Newell, S. J. & Goldsmith, R. E. (1997). The development of a scale to measure perceived
corporate credibility, Journal of Business Research, 52, 235-247. (N. Rifon, personal
communication, January 10, 2012).
O’Keefe, D.J. (2006). The advantages of compliance or the disadvantages of non-compliance?
A meta-analytic review of the relative persuasiveness of gain and loss-framed messages.
Communication Yearbook, 30, 1-43.
Owram, C. (3 September, 2008). As outbreak subsides, Maple Leaf must now rebuild consumer
confidence: experts. The Canadian Press
Patel, V. L., Gutnik, L.A., Yoskowitz, N.A., O'Sullivan, F. & Kaufman, D.R. (2006). Patterns of
reasons and decision making about condom use by urban college students, AIDS care,
18(8), 918-930.
Perdue, B.C. & Summers, J.O. (1986). Checking the success of manipulations in marketing
experiments, Journal of Marketing Research, 23, 317-326.

156

Prochaska, J.O. (1994). The transtheoretical approach: Crossing traditional boundaries of
therapy, Malabar, FL: Krieger.
Pyszczynski, T., Greenberg, J., & Solomon, S. (1997). Why do we need what we need? A terror
management perspective on the roots of human social motivation, Psychological Inquiry,
8, 1-20.
Retail Customer Experience (2012). At Kroger, shoppers are bananas for loyalty, retrieved from
http://www.retailcustomerexperience.com/blog_print/5756/At-Kroger.html
Reips, U.D. (1996). Experimenting in the World-Wide Web. Presented at the 26th annual
meeting of the Society of Computers in Psychology, Chicago, IL.
Rippetoe, P. & Rogers, R. (1987). Effects of Components of Protection Motivation Theory on
Adaptive and Maladaptive Coping with a Health Threat, Journal of Personality and
Social Psychology, 52, 596-604.
Robberson, M.R. & Rogers, R. (1988). Beyond fear appeals: Negative and positive persuasive
appeals to health and self-esteem, Journal of Applied Social Psychology, 18(3), 277-287.
Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change.
Journal of Psychology, 91, 93–114.
Rogers, R. (1983). Cognitive and Physiological Process in Fear Appeals and Attitude Change: A
Revised Theory of Protection Motivation, in Social Psychophysiology: A Source Book,
John Cacioppo and Richard Petty, eds. New York: Guilford Press, 153-176
Rogers, R., Deckner, C. & Mewborn, C. (1978). An expectancy-value theory approach to the
long-term modification of smoking behavior, Journal of Clinical Psychology, 34, 562566.
Rogers, R. & Mewborn, C. (1976). Fear appeals and attitude change: Effect of a threat's
noxiousness, probability of occurrence and the efficacy of coping appraisals, Journal of
Personality and Social Psychology, 34, 54-61.
Rogers, R. & Prentice-Dunn, S. (1997). Protection motivation theory. In D. Gochman (Ed.),
Handbook of health behavior research: Vol. 1 Personal and social determinants, 113-132.
New York: Plenum.
Rothman, A. J. & Salovey, P. (1997). Shaping perceptions to motivate healthy behavior: The
role of message framing, Psychological Bulletin, 121 (1), 3-19.
Rotter, J. (1966). Generalized expectancies for internal versus external control of reinforcement,
Psychological Monographs, 1, 1-28.

157

Rothman, A.J., Martino, S.C., Bedell, B.T., Detweiler, J.B., Salovey, P. (1999). The systematic
influence of gain and loss framed messages on interest in and use of different types of
health behavior, Personality and Social Psychology Bulletin, 121 (1), 3-19.
Sallam, M.A.A. (2011). The impact of source credibility on Saudi consumer’s attitude toward
print advertisement: The moderating role of brand familiarity, International Journal of
Marketing Studies, 3(4), 63-77.
Senior, C., Philips, M.L., Barnes, J., & David, A.S. (1999). An investigation in the perceptions
of dominance from schematic faces: A study using the World-Wide Web, Behavior
Research Methods, Instruments & Computers, 31, 341-346.
Settle, R.B. & Golden, L.L. (1974). Attribution theory and advertiser credibility, Journal of
Marketing Research, 11, 181-185.
Seydel, E., Taal, E., & Wiegman, O. (1990). Risk appraisal, outcome and self-efficacy
expectancies: Cognitive factors in previous behavior related to cancer. Psychology and
Health, 4, 99-109
Shaver, K.G. (1985). The attribution theory of blame: Causality, responsibility, and
blameworthiness, New York: Springer-Verlag
Sheehan, K. (2001). Online survey response rates: A review, Journal of Computer Mediated
Communication, 6 (2), retrieved from http://jcmc.indiana.edu/vol6/issue2/sheehan.html.
Shenk, D. (1997). Data smog: Surviving the information glut. New York: HarperEdge
Sherer, M., Maddux, J. E., Mercandante, B., Prentice-Dunn, S., Jacobs, B., & Rogers, R. (1982).
The self-efficacy scale: Construction and validation, Psychological Reports, 51, 663671.
Snow, D.A. & Benford, R.D. (1988). Ideology, frame resonance, and participant mobilization.
In B. Kandermans, H. Kriesi, & S. Tarrow (Eds.), International social movement
research. Vol. 1, From structure on action: Comparing social movement research across
cultures, Greenwich, CT: JAI Press, 197-217.
Stadden, S.A (2007). The influence of athletic identity, expectation of toughness, and attitude
toward pain and injury on athletes' help-seeking tendencies, (Doctoral dissertation).
Retrieved from ProQuest Dissertations and Theses Database.
Steffen, V. J. (1990). Men’s motivation to perform testicle self-exam: Effect of prior knowledge
and an educational brochure, Journal of Applied Social Psychology, 7, 101-113.
Survey System (2011). Sample Size Calculator. Available from:
http://www.surveysystem.com/sscalc.htm

158

Tanner, J. F., Hunt, J. B., & Eppright, D. R. (1991, July). The protection motivation model: A
normative model of fear appeals. Journal of Marketing, 55, 36–45.
Tversky, A. & Kahneman, D. (1981). The framing of decisions and the psychology of choice,
Science, 211 (4481), 453-458.
Tversky, A. & Kahneman, D. (1974). Judgment under uncertainty, heuristics and biases,
Science, 185 (4157), 1124-1131.
Tversky, A. & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation
of uncertainty, Journal of Risk and Uncertainty, 5(4), 297-323.
United States Census Bureau, 2011. The 2012 statistical abstract, Available from:
http://www.census.gov/compendia/statab/cats/population.html
United States Department of Agriculture, Food Safety Inspection Service (2008), Recall Case
Archive, retrieved from
http://www.fsis.usda.gov/Fsis_Recalls/Recall_Case_Archive/index.asp
Wagner, A. (2011). Bacterial Food Poisoning, retrieved from: http://aggiehorticulture.tamu.edu/extension/poison.html
Wilson, T. D., Kraft, D.S., & Lisle, D.J. (1989). Introspection, attitude change,, and attitude
behavior consistency: The disruptive effects of explaining why we feel the way we do,
Advances in Experimental Social Psychology, 22, 287-343.
Witte, K. (1994). Fear control and danger control: A test of the extended parallel process model.
Communication Monographs, 6, 113–13
Witte, K. & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective public
health campaigns. Health Education and Behavior, 27, 591-615.
Workman, M., Bommer, W. H., & Straub, D. (2008). Security lapses and the omission of
information security measures: A threat control model and empirical test, Computers in
Human Behavior, 24, 2799-2816.

159