Vous êtes sur la page 1sur 163

97

CHAPTER 6
EXPERIMENTAL STUDIES

6.1

OVERVIEW
This chapter provides a detailed description of the both the studies

and their associated results. Initially, the pretests are discussed, followed by a
description of the experiments in both the studies. Finally the results of the
experiments are examined and a discussion of the results is presented.
6.2

PRETESTS
Two pretests were conducted. The first pretest was done to identify

the products that were familiar and of interest to the study population. The
second pretest was done to identify the country that was considered proximal
to the target population. The second pretest was conducted to develop stimuli
for the experiments in Study 1. Both the pretests were conducted on forty four
MBA (first year) students in a large South Indian University. The average age
of the students was 20 and 56% of them were male. The questionnaire Q1 is
shown in Appendix 2.
6.2.1

Product Category
This pretest was conducted to identify products that were relevant

to the target population. A set of 10 products were selected for the pretest.
The selection was based on products used in extant green advertising studies
and popular products that used green advertisements in India (chosen from

98

Table 4.4 in Chapter 4). The ten products were laundry detergent (Schuhwerk
& Lefkoff-Hagius 1995; Kong & Zhang 2013), shampoo (Chang 2011),
mobile phone (Paladino & Ng 2013), mineral water (Grimmer & Woolley
2012), jeans, laptop, skin whiteners, scooter, notebooks and wristwatch.
Consumer involvement scale (Traylor & Joseph 1984) - a six item sevenpoint scale that is used to gauge consumers involvement across product
categories was used to measure consumer involvement with the selected
products (The scale Q1a is shown in Appendix 2). The results of the pretest
are shown in Table 6.1a. Mobile phones (M=17.45, S.D=7.949) and
wristwatches (M=15.93, S.D=5.699) were ranked high by the consumers.

Wristwatch

Notebook

Scooter

whitener

Skin

Bottle

Water

Laptop

Jeans

Mobile

Shampoo

Detergent

Table 6.1a Results of pretest for product preferrences

Mean

28.02 26.55 17.45 19.14 18.45 24.93

22.77

19.57 23.50 15.93

44

44

44

8.523

7.053 6.743 5.699

44

44

44

44

44

Std.
8.245 7.866 7.949 7.438 6.670 8.445
Deviation

44

44

A t-test was also conducted to verify if there was any relationship between
gender and product preferences. The results of the t-tests are shown in Table
6.1b. The results show that there were no gender differences in the product
preferences. Based on this pretest, mobile phones and wristwatches were
chosen as the products for the experiments.

99

Table 6.1b

Results of independent samples t-test to test gender


difference in product preferrences
Levene's
Test for
Equality of
Variances

Sig.

t-test for Equality of Means

df

95%
Confidence
Sig. *
Mean Std. Error Interval of
(2the
Difference Difference
tailed)
Difference
Lower Upper

Detergent

Equal
variances
assumed

.076

Equal
variances
not
assumed
Shampoo

Equal
variances
assumed

Jeans

Equal
variances
assumed
Equal
variances
not
assumed
Equal
variances
assumed

.725

.408

Equal
variances
not
assumed
Laptop

Equal
variances
assumed
Equal
variances
not
assumed

.813

.600

2.524

-4.494 5.694

.241 41.889 .811

.600

2.493

-4.431 5.631

.156

3.400

2.352

-1.347 8.147

1.475 41.946 .148

3.400

2.305

-1.251 8.051

.943

.175

2.435

-4.739 5.089

.074 40.911 .941

.175

2.359

-4.589 4.939

.741

.758

2.276

-3.834 5.351

.330 38.783 .743

.758

2.298

-3.890 5.406

.740

.633

1.899

-3.199 4.466

.326 35.889 .746

.633

1.940

-3.303 4.569

1.420 .240 1.446

Equal
variances
not
assumed
Mobile

.784 .238

.444

.399 .072

.527 .333

.509 .333

42

42

42

42

42

100

Table 6.1b (Continued)


Levene's
Test for
Equality of
Variances

Sig.

t-test for Equality of Means

df

95%
Confidence
Sig. *
Mean Std. Error Interval of
(2the
Difference Difference
tailed)
Difference
Lower Upper

Water_Bottle Equal
variances
assumed

.250

Equal
variances
not
assumed
Skin_whitener Equal
variances
assumed

.004

Equal
variances
not
assumed
Scooter

Equal
variances
assumed

Equal
variances
assumed

Equal
variances
assumed
Equal
variances
not
assumed

3.242

2.538

-1.881 8.364

1.284 41.320 .206

3.242

2.525

-1.856 8.339

.542

-1.600

2.599

-6.846 3.646

-.616 40.772 .541

-1.600

2.596

-6.844 3.644

.747

.700

2.158

-3.655 5.055

.339 38.753 .737

.700

2.068

-3.483 4.883

.507

-1.375

2.055

-5.522 2.772

-.682 41.976 .499

-1.375

2.015

-5.442 2.692

.335

1.683

1.727

-1.801 5.168

.963 38.160 .342

1.683

1.748

-1.855 5.221

.952 -.616

1.025 .317 -.669

Equal
variances
not
assumed
Wristwatch

.209

4.590 .038 .324

Equal
variances
not
assumed
Notebook

.619 1.277

.044

.835 .975

42

42

42

42

42

101

6.2.2

Geographical Proximity of Environmental Issues


Students rated the relevancy of environmental issues based on

geographical (spatial) proximity using seven point scales from relevant to


me to irrelevant to me (Chang 2012) (The scale Q1b is shown in
Appendix 2). Table 6.2 shows the results of this pretest.
It can be seen that issues related to South India (M=1.93
S.D=1.676) were considered highly relevant when compared to North India
and other countries (China, USA and Australia). Environmental issues in
Australia were considered least important (M=4.30 S.D=1.960).
Table 6.2 Pretest for geographical proximity
China North_India USA South_India Australia
Mean
N
Std. Deviation

6.3

STUDY

3.75

2.57

3.98

1.93

4.30

44

44

44

44

44

1.754

1.485

1.886

1.676

1.960

1:

EXPERIMENT

1:

TEMPORAL

AND

GEOGRAPHICAL FRAMING OF THREAT


This study was conducted to evaluate the effect of temporal and
geographical framing of threat on the PMT variables. The effect of the PMT
variables on involvement and the subsequent influence of involvement on
attitudes and purchase intention was also evaluated. The stimuli were
developed based on the products chosen using the pretests i.e mobile phone
and wristwatch. The experiment was also used to assess the stimuli, content
and face validity of the instrument.

102

6.3.1

Experimental Design
A 2 (temporal proximity of health threat: day vs. year) x 2

(geographical proximity of the health threat: local vs. global) between


subjects experimental design was utilized to investigate the hypotheses. This
resulted in four possible combinations of the factors. Fifty nine valid
responses were obtained from MBA students from a large South Indian
University (39 % male, median age=22). The students were randomly
assigned to the four possible conditions for the mobile phone stimuli.
Similarly, forty one valid responses were obtained from MBA students from a
large South Indian University (62 % male, median age=23). The students
were randomly assigned to the four possible conditions for the watch stimuli.
Data collection was through a paper and pencil questionnaire (Q2
shown in Appendix 3). Students first filled the questionnaire containing the
major dependent variables. Next, they were asked to answer a filler
questionnaire which asked them to describe their favourite celebrity. This was
a filler task designed to distract the respondents from associating the
personality variables question with the next questionnaire. On completion,
they filled counterbalanced questionnaires containing the questions for the
variables related to the environment (environmental concern, environmental
knowledge)

and

the

personality

variable

(consideration

for

future

consequences).
6.3.2

Stimuli
A total of four print advertisements were developed for the four

cells: temporally proximal threat and geographically proximal threat;


temporally proximal threat and geographically distant threat; temporally
distant threat and geographically proximal threat; temporally distant threat

103

and geographically distant threat. The advertisements also listed the


environment friendly features of the wristwatch or the mobile phone.
In the temporal threat conditions, the advertisement distinguished
between a day vs. year. Either day or year was used as the reference to specify
the number of people who suffer from respiratory diseases and cancers due to
toxins from either plastic waste (in case of watch) or electronic waste (in case
of mobiles) (Chandran & Menon 2004). In terms of geographical proximity
India was used to denote proximity and world was used to denote
geographically distant threats (Chang 2012). Although Australia was shown
as the location that is most geographically distant in the pretests, it was not
meaningful to represent an equivalent message that presented a threat in
Australia as part of the stimuli. Hence a more generic world was used to
denote a geographically distant threat. The ad was similar to those appearing
in the current Indian print media. The layout and format were not distinct and
contained basic information about the mobile phone or watch. The ad showed
a photograph of the mobile phone/watch and contained a description of the
product features and its environmental attributes. The mobile phone ad
contained a description of the display unit, talk time, standby time, OS and
memory. The mobile phone ad also specified that it contained recyclable
materials and avoided toxic components. The wristwatch ad contained details
about the materials and components used. The ad for the wristwatch also
highlighted its biodegradability.
6.3.3

Treatment Validity
The four print advertisements were analyzed by an expert panel to

assess if it contained the necessary variations in the temporal proximity and


the geographical proximity. This panel consisted of 3 marketing professors
and 2 Phd students who were familiar with marketing literature on fear

104

appeals, PMT and temporal framing. The panel suggested changes to the
presentation format and the final version of the advertisements are shown in
Appendix 4 (Figure A4.1, Figure A4.2, Figure A4.3, Figure A4.4, Figure
A4.5, Figure A4.6, Figure A4.7 and Figure A4.8)
6.3.4

Manipulation Checks
Manipulation checks were conducted by adding questions to verify

if the manipulations were successful. To this end, two questions were


included in the questionnaire. The temporal manipulation was checked asking
the question: How long do you think it takes for plastic waste pollution to
cause respiratory diseases or cancer?. The response to this item was
measured using seven point semantic scales anchored from 1 = the near future
and 7 = Distant future. Geographical manipulation was checked asking the
question: Is the issue of plastic pollution relevant to your country?. The
response to this item was measured using seven point semantic scales
anchored from 1 = Relevant to my country and 7 = Irrelevant to my country.
6.3.5

Dependent Variables
The study has mostly used previously validated instruments to

measure the constructs. The dependent variables include perceived severity,


perceived vulnerability, perceived self-efficacy, perceived response-efficacy,
fear, message involvement, attitude towards the advertisement, attitude
towards the brand and purchase intention. The other variables related to
individual

characteristics

included

environmental

concern,

objective

environmental knowledge and consideration for future consequences. The


sources and scales are described in detail below and shown in Appendix 3 as
discussed previously.

105

6.3.5.1

Protection motivation theory variables


The PMT variables were adapted from Milne et al (2002). They are

described in detail below:


Perceived severity
Perceived severity was measured using a four item seven point
scale where 1 = Strongly Disagree and 7 = Strongly Agree. Participants were
asked to indicate their responses on the following statements: I believe that
plastic waste / e-waste in the environment may cause severe health issues like
respiratory diseases and cancer, I believe that plastic waste / e-waste
pollution is a serious threat to human health, I do not think that plastic
waste /e-waste will affect our health, I believe plastic waste / e-waste
pollution is a significant problem. In the case of wristwatch plastic waste
was used and e-waste was used with mobile phone stimuli.
Perceived vulnerability
Perceived vulnerability was measured using a three item seven
point scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine
the participants susceptibility to the threat. Participants were asked to
indicate their responses to the following statements: I am worried that I
might get respiratory illness or cancer because of plastic waste / e-waste,
Plastic waste /E-waste pollution is a big concern for me as it might affect my
health, It is possible that I am at risk of being affected by respiratory illness
or cancer because of plastic waste / e-waste.
Response efficacy
Response efficacy was also measured using a three item seven
point scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine

106

whether participants believed if purchasing recyclable / biodegradable


products averted the threat. Participants rated their responses on the following
statements: Buying biodegradable / recyclable products is highly effective in
preventing diseases due to plastic/e-waste pollution, Buying biodegradable /
recyclable products will significantly lower the risk of being affected by
respiratory diseases and cancer, Buying biodegradable products is an
effective method of reducing respiratory illness and cancer in humans. In the
case of mobile phone, the word biodegradable was changed to recyclable.
Self efficacy
Self efficacy was measured using a three item seven point scale
where 1 = Strongly Disagree and 7 = Strongly Agree to determine whether
participants believed if they were capable of averting the threat. Participants
rated their responses to the following statements: It would be easy for me to
identify a biodegradable watch/ mobile made of recycled materials, It is not
difficult for me to check if the watch contains plastic or not / mobile is made
of recycled materials or not, I can easily identify a biodegradable watch /
mobile made of recycled materials.
Fear
Fear is an affective response to the threat levels presented in the
stimuli. Participants rated their emotions (the extent to which they
experienced each of the emotions afraid, scared, fearful, anxious and worried)
while viewing the advertisement on a seven item seven point Likert Scale (1 =
Strongly Disagree and 7 = Strongly Agree). This measure is similar to fear
measures used in previous studies that employ the protection motivation
theory and the scale had a high internal reliability (Hartmann et al 2013).

107

6.3.5.2

Message involvement
Participants reported agreement with six statements (on a seven

point Likert scale) adapted from Cox & Cox (2001) : I got involved in what
the advertisement had to say, The ad's message seemed relevant to me,
This ad really made me think, This ad was thought-provoking , The ad
was very interesting, and I felt strong emotions while reading this ad. This
scale had a good internal reliability score in previous studies (Cox & Cox
2001; Cauberghe et al 2009).
6.3.5.3

Attitudes and Intentions

Attitude towards ad
Attitude towards the ad was measured by using three seven point
semantic

differential

scales:

good/bad,

pleasant/unpleasant,

and

favorable/unfavorable ( =0.88) (Mackenzie & Lutz 1989).


Attitude towards the brand
Attitude towards the brand was measured by using three seven
point semantic differential scales: good/bad, pleasant/unpleasant, and
favorable/unfavorable ( =0.93) (Muehling & Laczniak 1988).
Purchase intention
Participants were asked to respond to three sets of bipolar
adjectives (unlikely-likely, definitely would-definitely would not, improbableprobable) placed on seven point scales to indicate how likely they were to
purchase the advertised brand. This scale was also adapted from previous
research (MacKenzie et al 1986).

108

6.3.5.4

Variables related to the environment and personality

Environmental concern
Participants environmental concern was measured using the scale
proposed by Schultz (2001). The scale requires the participants to rank their
environmental concerns from one to seven on sub-categories namely
biospheric concerns (plants, marine life, birds, and animals), altruistic
concerns (humanity, children, people in the country, future generations) and
egoistic concerns (me, my future, my health, my lifestyle).
Objective environmental knowledge
Objective environmental knowledge was measured using a set of
fifteen questions similar to the MEAK subscale on environmental knowledge
(Maloney et al 1975). The questions were based on combination of general
questions about environmental awareness (for instance, impact of climate
change, pollutants in batteries and CFLs) and issues specific to India (for
example, Bhopal disaster, maximum greenhouse emissions in India). Some of
the questions were taken from an online quiz (http://edugreen.teri.res.in/
explore/quiz/quiz.htm). The scale is in a quiz format and the correct answers
are summed to form the objective environmental knowledge score. Higher
scores reveal a high degree of factual knowledge about the environment and
vice-versa.
Consideration for future consequences (CFC)
Individuals temporal orientation was measured using the
consideration of future consequences fourteen item scale (Joireman et al
2012). The scale has two components measuring the concern with future

109

consequences and concern with immediate consequences. The consideration


for future consequences score was determined after recoding the immediate
items (3, 4, 5, 9, 10, 11, 12).
6.3.6

Results of Experiment 1
The experiment was conducted with the mobile phone and

wristwatch stimuli to evaluate the effect of temporal and geographical


framing of threat on the PMT variables. The effect of PMT variables on
involvement and the subsequent influence of involvement on attitudes and
intention was also evaluated. The experiment was also used to assess the
content and face validity of the instrument. The results are discussed below.
6.3.6.1

Manipulation check

Mobile phone stimuli


Temporal proximity manipulations were not successful as there was
no difference in the way participants evaluated the temporal proximity of the
threat. There was no significant main effect of the temporal proximity
manipulation (day vs. year) in both the proximal (M = 3.66) and distal (M =
3.87) conditions (F (1,57) = 0.224 ; p>0.5). The mean values and the ANOVA
tests are shown in Table 6.3a and Table 6.3b.
Table 6.3a shows that the mean values are very close in value. It
can also be seen from Table 6.3b that the temporal proximity did not have any
effect on the manipulation check variable. Geographical manipulations were
checked next.

110

Table 6.3a Experiment 1: Mean values of the manipulation check


variable for temporal distance of threat (mobile phone
stimuli)

Day

MC_TIME
95%
Confidence
Interval for
Std.
Std.
N Mean
Minimum Maximum
Mean
Deviation Error
Lower Upper
Bound Bound
29 3.66
1.610
.299
3.04
4.27
1
6

Year 30 3.87
Total 59 3.76

1.814
1.705

.331
.222

3.19
3.32

4.54
4.21

1
1

7
7

Table 6.3b Experiment 1: One way ANOVA - manipulation check


variable for temporal distance of threat (mobile phone
stimuli)
ANOVA
MC_TIME
Sum of Squares df Mean Square F
Sig.
Between Groups
.660
1
.660
.224 .638
Within Groups
Total

168.018
168.678

57
58

2.948

Similarly geographical proximity manipulations were also not successful as


there was no significant main effect of the geographical proximity
manipulations as proximity manipulation (India vs. World) did not produce
any statistically significant effect in both the proximal (M = 2.37) and distal
(M = 1.90) conditions (F (1,57) = 2.552 ; p>0.05). The results of the
geographical manipulation checls are shown in Tables 6.4a and 6.4b.

111

Table 6.4a

Experiment 1: Mean value of the manipulation check


variable for geographical distance of threat (mobile phone
stimuli)
Descriptives

MC_GEOGRAPHY
95% Confidence
Interval
for Mean
Std.
Std.
N Mean
Minimum Maximum
Deviation Error Lower
Upper
Bound
Bound
India 30 2.37 1.299
.237
1.88
2.85
1
6
World 29 1.90
Total 59 2.14

Table 6.4b

.939
1.152

.174
.150

1.54
1.84

2.25
2.44

1
1

4
6

Experiment 1: One way ANOVA of the manipulation check


variable for geographical distance of threat (mobile phone
stimuli)
ANOVA

MC_GEOGRAPHY

Between Groups

Sum of
Squares
3.259

Mean
Square
3.259
1.292

df

Within Groups

73.656

57

Total

76.915

58

Sig.

2.522

.118

Watch stimuli
Temporal proximity manipulations were not successful for the
watch stimuli as there was no significant main effect of the temporal
proximity manipulation (day vs. year) in both the proximal (M = 4.05) and
distal (M = 3.55) conditions (F (1, 40) = 0.990; p>0.05). The results are
shown in Tables 6.5a, 6.5b.

112

Table 6.5a Experiment 1: Mean values of the manipulation check


variable for temporal distance of threat (watch stimuli)
Descriptives
MC_TIME
Std.
Deviation

Std.
Error

Day 22 4.05
Year 20 3.55

1.704
1.504

Total 42 3.81

1.612

N Mean

95% Confidence
Interval for Mean

.363
.336

Lower
Bound
3.29
2.85

Upper
Bound
4.80
4.25

.249

3.31

4.31

Minimum Maximum

1
1

7
7

Table 6.5b Experiment 1: One way ANOVA of the manipulation check


variable for temporal distance of threat (watch stimuli)
ANOVA
MC_TIME
Sum of
Squares

df

Mean
Square

Sig.

2.572

2.572

.990

.326

Within Groups

103.905

40

2.598

Total

106.476

41

Between Groups

Similarly geographical proximity manipulations were also not successful as


there was no significant main effect of the geographical proximity
manipulations (India vs. World) in both the proximal (M = 1.96) and distal
(M = 2.24) conditions (F(1,40) = 0.414 ; p>0.05). Tables 6.5c and 6.5d show
these results.

113

Table 6.5c Experiment 1: Mean values of the manipulation check


variable for geographical distance of threat (watch stimuli)
Descriptives
MC_GEOGRAPHY

N Mean

Std.
Std.
Deviation Error

95% Confidence
Interval for Mean
Lower
Bound

Upper
Bound

Minimum Maximum

India 25 1.96

1.306

.261

1.42

2.50

World 17 2.24

1.437

.349

1.50

2.97

Total 42 2.07

1.351

.208

1.65

2.49

Table 6.5d Experiment 1: One way ANOVA of the manipulation check


variable for geographical distance of threat (watch stimuli)
ANOVA
MC_GEOGRAPHY
Sum of
Squares

df

Mean
Square

Sig.

.767

.767

.414

.523

Within Groups

74.019

40

1.850

Total

74.786

41

Between Groups

A failed manipulation check in social psychology research is not of


great concern and does not indicate that the manipulation of the independent
variable failed (Sigall & Mills 1998). Therefore further analyses on the data
were conducted.

114

6.3.6.2

Scale reliability

Mobile phone stimuli


The internal consistency of the scales was assessed using Cronbach
. Table 6.6 below shows the reliability scores. Almost all the constructs
meet and exceed 0.6 the rule of thumb criteria suggested by Nunnally
(Nunnally 1970) indicating that the instrument is reasonably reliable.
Although self-efficacy has a lower reliability score, most PMT studies report
such low score. Since

> 0.5 is acceptable, the same measure was used and

the reliabilities are deemed acceptable. As this study was also used to evaluate
the scales, the scale was retained. Environmental knowledge is treated as a
single formative indicator and therefore reliability score was not calculated
for this measure as it is illogical to check correlations between the indicators
for such a construct (Chin 1998).
Table 6.6 Experiment 1: Reliability scores using mobile phone stimulus
Construct
Cronbach
Perceived severity
0.64
Perceived vulnerability
0.88
Response Efficacy
0.82
Self Efficacy
0.52
Message involvement
0.82
Fear
0.89
Attitude towards ad
0.81
Attitude towards brand
0.90
Purchase intention
0.96
Environmental Knowledge
Environmental concern
0.86
Consideration for future consequences
0.85

115

Watch stimuli
Table 6.7 below shows the reliability scores for the watch stimuli.
Table 6.7 Experiment 1: Reliability scores using watch stimulus
Construct

Cronbach

Perceived severity

0.57

Perceived vulnerability

0.90

Response Efficacy

0.88

Self Efficacy

0.41

Message involvement

0.82

Fear

0.89

Attitude towards ad

0.82

Attitude towards brand

0.94

Purchase intention

0.93

Environmental concern

0.89

Consideration for future consequences

0.78

It can be seen from Table 6.7 that almost all the constructs meet
and exceed 0.6 Nunnallys rule of thumb (1970). The instrument is therefore
reasonable reliable. Self-efficacy has a lower reliability score and

< 0.5 is

unacceptable. Therefore, this measure was not used for further analysis with
the watch stimulus.
6.3.6.3

Hypotheses tests of the effect of manipulations on PMT


variables
The

hypothesized

effect

of

temporal

and

geographical

manipulations on the PMT variables was analyzed using MANOVA or


MANCOVA as appropriate. After checking for missing data and outliers, it

116

was found that most of the PMT variables were negatively skewed.
MANOVA is robust to violations of multivariate normality and to violations
of homogeneity of variance/covariance matrices if groups are of nearly equal
size (Leech et al 2011). The dependent variables (perceived severity,
perceived vulnerability, fear, response-efficacy and self-efficacy) were
moderately correlated (0.27 0.49) and therefore there was no risk of
multicollinearity to pose a hindrance to conductiong MANOVA.
Mobile phone stimuli
Table 6.8a shows the distribution characteristics of the protection
motivation variables and Table 6.8b shows the group wise means. It can be
seen that most of the variables have a mean value that is closer to the highest
score on the scale i.e. 7. Perceived severity ranks high among the threat
appraisal variables with a mean value of 5.91.
Table 6.8a

Experiment 1: Distribution characteristics of the protection


motivation variables (mobile phone stimuli)

Variable

Minimum Maximum Mean Std. Deviation

PERC_SEV

4.00

7.00

5.91

0.73

PERC_VUL

1.00

7.00

4.68

1.22

RESP_EFFICACY

2.00

7.00

4.98

1.25

SELF_EFFICACY

3.50

7.00

4.41

1.72

FEAR

2.00

6.14

3.60

0.92

Table 6.8b below does not show much variation across the groups
either. The perceived severity and perceived vulnerability scores appear close
in almost all the conditions.

117

Table 6.8b

Experiment 1: Group wise mean values of protection


motivation variables for the mobile phone stimuli
Self

Perceived
Severity

Perceived
Vulnerability

Response
Efficacy

Efficacy

Temporal
proximity: Day

6.12

4.93

4.82

4.31

3.70

Temporal
proximity: Year

5.72

4.45

5.13

4.52

3.49

Geographical
proximity: India

5.83

4.50

5.13

4.74

3.53

Geographical
proximity: World

6.00

4.88

4.82

4.08

3.67

Factor

Fear

Hypothesis 1 stated that consumers who viewed advertisements that


contained threats proximal in time would perceive higher severity and high
vulnerability when compared to consumers who viewed threats that were
distant in time. A one-way MANOVA was conducted to ascertain if there
were significant differences regarding perceived severity and perceived
vulnerability, among the groups in response to manipulation of the temporal
proximity of the threat. Cell sizes were approximately equal (29 and 30) and
the Box's Test indicated that the assumptions of normality were not violated
as there was no significant differences between the covariance matrices. The
one-way MANOVA results were: Pillais Trace=0.078; Wilks lambda =
0.992; Hotellings Trace and Roys Largest Root = 0.085, F(2,56)=2.379 as
shown in Table 6.9a. Since the results of the multivariate tests were not
significant, the dependent variables are not significantly dependent on the
temporal proximity of the threat. The results indicate that there was no
statistically significant difference in severity or vulnerability based on
temporal proximity. Therefore hypothesis 1 (H1) was not supported. Table
6.9a and 6.9b shows the results of the test. Although the multivariate tests
were not significant, Table 6.9b showed the possibility of an influence of the
time factor on perceived severity of the threat.

118

Table 6.9a

Experiment1: Hypothesis 1: multivariate tests (mobile


phone stimuli)
Multivariate Testsb
Effect

Intercept

Value

Pillai's Trace
.986 1980.986a
Wilks' Lambda
.014 1980.986a
Hotelling's Trace 70.749 1980.986a
Roy's Largest Root 70.749 1980.986a
Multivariate Testsb
Effect

Value

TIME_FACTOR Pillai's Trace


.078
Wilks' Lambda
.922
Hotelling's Trace
.085
Roy's Largest Root .085
a. Exact statistic
b. Design: Intercept + TIME_FACTOR

Table 6.9b

F
2.379a
2.379a
2.379a
2.379a

Hypothesis
df
2.000
2.000
2.000
2.000

Error
df
56.000
56.000
56.000
56.000

Hypothesis
df
2.000
2.000
2.000
2.000

Error
df
56.000
56.000
56.000
56.000

Sig.
.000
.000
.000
.000
Sig.
.102
.102
.102
.102

Experiment 1: Hypothesis 1: tests of between-subjects


effects (mobile phone stimuli)

Tests of Between-Subjects Effects


Type III
Dependent
Mean
Source
Sum of
df
Variable
Square
Squares
Corrected Model PERC_SEV
2.309a
1 2.309
PERC_VUL
3.334 b
1 3.334
Intercept
PERC_SEV
2069.131
1 2069.131
PERC_VUL
1299.221
1 1299.221
TIME_FACTOR PERC_SEV
2.309
1 2.309
PERC_VUL
3.334
1 3.334
Error
PERC_SEV
29.246
57 .513
PERC_VUL
84.192
57 1.477
Total
PERC_SEV
2098.938
59
PERC_VUL
1384.889
59
Corrected Total
PERC_SEV
31.555
58
PERC_VUL
87.525
58
a. R Squared = .073 (Adjusted R Squared = .057)
b. R Squared = .038 (Adjusted R Squared = .021)

Sig.

4.500
2.257
4032.657
879.607
4.500
2.257

.038
.139
.000
.000
.038
.139

119

A follow up one-way ANOVA was hence conducted to test if the


temporal proximity of the threat had an effect on the perceived severity.
Table 6.9c shows the results of the test. There was a statistically significant
difference

between

groups

(F(1,57) = 4.500, p <.05).

as

determined

by

one-way

ANOVA

It can be seen from Table 6.9c that temporal

proximity of the threat influenced the perceived severity of the threat. A plot
was producted to check the effect. Figure 6.1 shows that participants who
viewed threats that were closer in time perceived higher levels of severity.
This implies that the temporal proximity of the threat has an effect on the
perceived severity.
Table 6.9c Experiment 1: Tests of between-subjects effects of the
temporal proximity of threat on perceived severity (mobile
phone stimuli)
Tests of Between-Subjects Effects
Dependent Variable:PERC_SEV

Source

Type III
Sum of df
Squares

Corrected Model
Intercept

2309a

2.309

4.500

Squared

Noncent Observed
Parameter Powerb

.073

4.500

.550

2069.131 1 2069.131 4032.657 .000

.986

4032.657

1.000

.073

4.500

.550

2.309

Error

29.246 57

.513

Total

2098.938 59

CorrectedTotal

Partial
Sig
Eta
.038

TIME_FACTOR 2.309

Mean
Square

4.500

31.555 58

a.RSquared=.073 (Adjusted R Squared = .057)


b. Computed using alpha = .05

.038

120

Figure 6.1

Experiment 1 effect of framing a temporally and


geographically proximal threat on perceived severity

The main effect of geographical proximity of threat on perceived


severity and vulnerability to the threat was investigated using one-way
MANOVA. Hypothesis 2 (H2) was also not supported as there was no
statistically significant difference in severity and vulnerability based on the
geographical proximity of threat. The one-way MANOVA results were:
Pillais Trace=0.027; Wilks lambda = 0.876; Hotellings Trace and Roys
Largest Root = 0.028, F(2,56)=0.787. The results are shown in Table 6.10a
and 6.10b.

121

Table 6.10a Experiment 1: Hypothesis 2: multivariate tests (mobile


phone stimuli)
Multivariate Tests b
Effect
Intercept

Value

Hypothesis df Error df Sig.

Pillai's Trace

.985 1861.603a

2.000

56.000 .000

Wilks' Lambda

.015 1861.603a

2.000

56.000 .000

66.486 1861.603a

2.000

56.000 .000

Roy's Largest Root 66.486 1861.603a

2.000

56.000 .000

Hotelling's Trace

.027

.787a

2.000

56.000 .460

Wilks' Lambda

.973

.787a

2.000

56.000 .460

Hotelling's Trace

.028

.787a

2.000

56.000 .460

Roy's Largest Root .028

.787a

2.000

56.000 .460

GEOG_FACTOR Pillai's Trace

a. Exact statistic
b. Design: Intercept + GEOG_FACTOR

Table 6.10b Experiment 1: Hypothesis 2: tests of between-subjects


effects (mobile phone stimuli)
Tests of Between-Subjects Effects
Type III
Sum of
Squares

df

Mean
Square

Sig.

Corrected Model PERC_SEV

.453a

.453

.830

.366

PERC_VUL

2.186b

2.186

1.460

.232

PERC_SEV

2067.826

2067.826 3789.661 .000

PERC_VUL

1298.797

1298.797 867.497

GEOG_FACTOR PERC_SEV

.453

.453

.830

.366

PERC_VUL

2.186

2.186

1.460

.232

PERC_SEV

31.102

57

.546

PERC_VUL

85.339

57

1.497

Source

Intercept

Error

Dependent
Variable

.000

122

Table 6.10b (Continued)


Source

Dependent

Type III

Variable

Sum of

df

Mean

Sig.

Square

Squares
Total

Corrected Total

PERC_SEV

2098.938

59

PERC_VUL

1384.889

59

PERC_SEV

31.555

58

PERC_VUL

87.525

58

a. R Squared = .014 (Adjusted R Squared = -.003)


b. R Squared = .025 (Adjusted R Squared = .008)

Hypothesis 3 predicted interaction effects and stated that an


interaction between temporal proximity and geographical proximity would
cause perception of higher levels of severity and vulnerability under proximal
conditions. A 2 (temporal proximity of threat:day vs.year) x 2 (geographical
proximity of threat: India vs. World)

multivariate analysis of variance

(MANOVA) was conducted to examine this. The results also indicated that no
interaction effect exists between temporal proximity of threat and
geographical proximity of threat on both the PMT variables (Pillais
Trace=0.016; Wilks lambda = 0.984; Hotellings Trace and Roys Largest
Root = 0.017, F(2,54)=0.450). The results are shown in Table 6.11a and Table
6.11b. Therefore, Hypothesis 3 was not accepted.

123

Table 6.11a Experiment 1: Hypothesis 3: multivariate tests (mobile


phone stimuli)
Multivariate Tests b
Effect
Intercept

Value

GEOG_FACTOR

TIME_FACTOR *
GEOG_FACTOR

Hypothesis Error
Sig.
df
df

Pillai's Trace

.986 1972.206a

2.000

54.000 .000

Wilks'
Lambda

.014 1972.206a

2.000

54.000 .000

73.045 1972.206a

2.000

54.000 .000

Roy's
73.045 1972.206a
Largest Root

2.000

54.000 .000

Hotelling's
Trace

TIME_FACTOR

Pillai's Trace

.082

2.398a

2.000

54.000 .101

Wilks'
Lambda

.918

2.398a

2.000

54.000 .101

Hotelling's
Trace

.089

2.398a

2.000

54.000 .101

Roy's
Largest Root

.089

2.398a

2.000

54.000 .101

Pillai's Trace

.030

.846a

2.000

54.000 .435

Wilks'
Lambda

.970

.846a

2.000

54.000 .435

Hotelling's
Trace

.031

.846a

2.000

54.000 .435

Roy's
Largest Root

.031

.846a

2.000

54.000 .435

Pillai's Trace

.016

.450a

2.000

54.000 .640

Wilks'
Lambda

.984

.450a

2.000

54.000 .640

Hotelling's
Trace

.017

.450a

2.000

54.000 .640

Roy's
Largest Root

.017

.450a

2.000

54.000 .640

a. Exact statistic
b. Design: Intercept + TIME_FACTOR + GEOG_FACTOR + TIME_FACTOR *
GEOG_FACTOR

124

Table 6.11b Experiment 1: Hypothesis 3: tests of between-subjects


effects (mobile phone stimuli)
Tests of Between-Subjects Effects
Dependent
Variable

Type III
Sum of
Squares

df

Mean
Square

Sig.

PERC_SEV

3.228a

1.076

2.089

.112

PERC_VUL

6.267

2.089

1.414

.249

PERC_SEV

2067.929

2067.929

4015.101 .000

PERC_VUL

1299.290

1299.290

879.425

.000

PERC_SEV

2.309

2.309

4.483

.039

PERC_VUL

3.375

3.375

2.284

.136

PERC_SEV

.473

.473

.919

.342

PERC_VUL

2.238

2.238

1.515

.224

TIME_FACTOR *
GEOG_FACTOR

PERC_SEV

.430

.430

.835

.365

PERC_VUL

.652

.652

.441

.509

Error

PERC_SEV

28.327

55

.515

PERC_VUL

81.259

55

1.477

PERC_SEV

2098.938

59

PERC_VUL

1384.889

59

PERC_SEV

31.555

58

PERC_VUL

87.525

58

Source
Corrected Model
Intercept
TIME_FACTOR
GEOG_FACTOR

Total
Corrected Total

a. R Squared = .102 (Adjusted R Squared = .053)


b. R Squared = .072 (Adjusted R Squared = .021)

To test H4, a one way MANCOVA with perceived severity and


perceived vulnerability as dependent variable and CFC as the covariate was
conducted. The temporal proximity of the threat was the independent variable.
The assumptions for MANCOVA were met. In particular, the homogeneity of
the regression effect was evident for the covariate, and the covariate was
linearly related to the dependent measure. The one-way MANCOVA results
were as follows: (Pillais Trace=0.064; Wilks lambda = 0.936; Hotellings
Trace and Roys Largest Root = 0.068, F(2,54)=1.846). There were no

125

interaction effects and therefore H4 was not supported. The results are shown
in Table 6.12a and 6.12b below. It can be seen from Table 6.12a that the
temporal proximity of the threat did not have a significant effect on the
hypothesized variables. Hence Table 6.12b was not interpreted.
Table 6.12a Experiment 1: Hypothesis 4: multivariate tests (mobile
phone stimuli)
Multivariate Tests b
Effect
Intercept

TIME_FACTOR

TIME_FACTOR *
CFC_TOTAL

CFC_TOTAL

Pillai's Trace
Wilks'
Lambda
Hotelling's
Trace
Roy's Largest
Root
Pillai's Trace
Wilks'
Lambda
Hotelling's
Trace
Roy's Largest
Root
Pillai's Trace
Wilks'
Lambda
Hotelling's
Trace
Roy's Largest
Root
Pillai's Trace
Wilks'
Lambda
Hotelling's
Trace
Roy's Largest
Root

87.364a
87.364a

Hypothesis
df
2.000
2.000

Error
df
54.000
54.000

.000
.000

3.236

87.364a

2.000

54.000

.000

3.236

87.364a

2.000

54.000

.000

.087
.913

2.585a
2.585a

2.000
2.000

54.000
54.000

.085
.085

.096

2.585a

2.000

54.000

.085

.096

2.585a

2.000

54.000

.085

.064
.936

1.846a
1.846a

2.000
2.000

54.000
54.000

.168
.168

.068

1.846a

2.000

54.000

.168

.068

1.846a

2.000

54.000

.168

.097
.903

2.902a
2.902a

2.000
2.000

54.000
54.000

.063
.063

.107

2.902a

2.000

54.000

.063

.107

2.902a

2.000

54.000

.063

Value

.764
.236

a. Exact statistic
b. Design: Intercept + TIME_FACTOR + TIME_FACTOR * CFC_TOTAL +
CFC_TOTAL

Sig.

126

Table 6.12b Experiment 1: Hypothesis 4: tests of between-subjects effects


Tests of Between-Subjects Effects
Type III
Sum of
Squares

df

Mean
Square

Sig.

PERC_SEV

5.611 a

1.870

3.965

.012

PERC_VUL

15.982b

5.327

4.096

.011

PERC_SEV

83.946

83.946

177.959 .000

PERC_VUL

34.893

34.893

26.824

.000

PERC_SEV

1.585

1.585

3.360

.072

PERC_VUL

5.090

5.090

3.913

.053

TIME_FACTOR *
CFC_TOTAL

PERC_SEV

.984

.984

2.086

.154

PERC_VUL

3.981

3.981

3.060

.086

CFC_TOTAL

PERC_SEV

1.640

1.640

3.476

.068

PERC_VUL

6.052

6.052

4.653

.035

PERC_SEV

25.944

55

.472

PERC_VUL

71.543

55

1.301

PERC_SEV

2098.938

59

PERC_VUL

1384.889

59

PERC_SEV

31.555

58

PERC_VUL

87.525

58

Source
Corrected Model

Intercept
TIME_FACTOR

Error
Total
Corrected Total

Dependent
Variable

a. R Squared = .178 (Adjusted R Squared = .133)


b. R Squared = .183 (Adjusted R Squared = .138)

Since the hypothesis testing did not yield any specific results, a
three way ANOVA was conducted to examine the interactions among the
factors and covariates on the individual dependent variables (perceived
severity and perceived vulnerability). Tables 6.13a and 6.13b show the result.
The results showed significant interactions of the factors and the covariate
CFC on perceived severity and perceived vulnerability of the threat.

127

Table 6.13a Experiment 1: Tests of between-subjects effects (effect of


temporal, geographical proximity of threat and CFC on
perceived severity)
Tests of Between-Subjects Effects
Dependent Variable:PERC_SEV
Type III
Source

Sum of df
Squares

Mean
Square

Partial
F

Sig.

Eta
Squared

Noncent. Observed
Parameter Powerb

Corrected Model

9.538a

7 1.363

3.156 .008

.302

22.093

.918

Intercept

76.799

1 76.799 177.895 .000

.777

177.895

1.000

TIME_FACTOR

3.449

1 3.449

7.988 .007

.135

7.988

.792

GEOG_FACTOR

2.536

1 2.536

5.875 .019

.103

5.875

.662

CFC_TOTAL

.209

.209

.485 .489

.009

.485

.105

2.492

1 2.492

5.773 .020

.102

5.773

.654

2.093

1 2.093

4.849 .032

.087

4.849

.579

1.875

1 1.875

4.343 .042

.078

4.343

.534

1.545

1 1.545

3.579 .064

.066

3.579

.459

TIME_FACTOR *
CFC_TOTAL
GEOG_FACTOR
* CFC_TOTAL
TIME_FACTOR *
GEOG_FACTOR
* CFC_TOTAL
TIME_FACTOR *
GEOG_FACTOR
Error
Total
Corrected Total

22.017 51 .432
2098.938 59
31.555 58

a. R Squared = .302 (Adjusted R Squared = .206)


b. Computed using alpha = .05

128

Table 6.13b Experiment 1: Tests of between-subjects effects (effect of


temporal, geographical proximity of threat and cfc on
perceived vulnerability)
Tests of Between-Subjects Effects
Dependent Variable:PERC_VUL
Source

Type III
Mean
Sum of df
Square
Squares

Partial
Noncent. Observed
Sig. Eta
Parameter Powerb
Squared

Corrected Model

28.190a

7 4.027 3.461 .004

.322

24.230

.943

Intercept

43.236

1 43.236 37.162 .000

.422

37.162

1.000

TIME_FACTOR

10.796

1 10.796 9.279 .004

.154

9.279

.848

GEOG_FACTOR

9.775

1 9.775 8.402 .006

.141

8.402

.812

CFC_TOTAL

.783

.673 .416

.013

.673

.127

8.870

1 8.870 7.624 .008

.130

7.624

.773

GEOG_FACTOR
* CFC_TOTAL

8.071

1 8.071 6.937 .011

.120

6.937

.734

TIME_FACTOR *
GEOG_FACTOR
* CFC_TOTAL

3.782

1 3.782 3.251 .077

.060

3.251

.424

TIME_FACTOR *
GEOG_FACTOR

3.343

1 3.343 2.874 .096

.053

2.874

.384

Error

59.336 51 1.163

TIME_FACTOR *
CFC_TOTAL

Total

.783

1384.889 59

Corrected Total

87.525 58

a. R Squared = .322 (Adjusted R Squared = .229)


b. Computed using alpha = .05

The following Table (Table 6.14) summarizes the results of the


hypotheses testing for the effect of the stimuli on the PMT variables (H1-H4).

129

Table 6.14

Experiment 1: Summary of the hypotheses (H1 H4)


mobile phone stimulus

Hypothesis

Factor

Perceived

Perceived

severity

vulnerability

H1

Temporal proximity (H1)

H2

Geographical proximity (H2)

H3

Temporal proximity *

Geographical proximity (H3)


H4

CFC (H4)

X no effect

- Effect present

Watch stimuli
Table 6.15 shows the distribution characteristics of the protection
motivation variables. Similar to the mobile phone stimulus, most of the values
were negatively skewed. Group averages for the variables are shown in Table
6.16. Self efficacy was not included in the analysis as the scale reliability was
very low.
Table 6.15

Experiment 1: Distribution characteristics of the protection


motivation variables for the watch stimuli

Variable

Minimum Maximum Mean Std. Deviation

Perceived Severity

2.00

7.00

6.39

0.92

Perceived Vulnerability

1.00

7.00

4.73

1.43

Response Efficacy

2.33

7.00

5.15

1.24

Fear

1.83

6.50

3.86

0.99

130

Table 6.16 Experiment 1: Group wise mean values of protection


motivation variables for the watch stimuli
Self
Fear
Efficacy

Perceived
Severity

Perceived
Vulnerability

Response
Efficacy

6.27

4.60

5.13

3.87

6.52

4.86

5.16

3.84

Geographical
proximity: India

6.60

4.64

5.18

3.80

Geographical
proximity: World

6.08

4.86

5.09

3.94

Factor
Temporal
proximity: Day
Temporal
proximity: Year

A one-way MANOVA was conducted to verify H1 which stated


that there will be significant differences in perceived severity and perceived
vulnerability among the groups in response to manipulation of the temporal
proximity of the threat. H1 was not accepted (Pillais Trace=0.021; Wilks
lambda = 0.979; Hotellings Trace and Roys Largest Root = 0.021,
F(2,39)=0.415. Table 6.17a and 6.17b show the results.
Table 6.17a Experiment 1: Hypothesis 1: multivariate tests (watch
stimuli)

Effect
Intercept

Multivariate Testsb
Value F
Hypothesis df Error df Sig.
a
Pillai's Trace
.981 989.360
2.000
39.000 .000
.019 989.360a
50.736 989.360a

2.000
2.000

39.000 .000
39.000 .000

Roy's Largest Root 50.736 989.360a

2.000

39.000 .000

2.000
2.000
2.000
2.000

39.000
39.000
39.000
39.000

Wilks' Lambda
Hotelling's Trace
TIME_FACTOR Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root

.021
.979
.021
.021

a. Exact statistic
b. Design: Intercept + TIME_FACTOR

.415
.415a
.415 a
.415a

.664
.664
.664
.664

131

Table 6.17b Experiment 1: Hypothesis 1: tests of between-subjects


effects (watch stimuli)
Tests of Between-Subjects Effects

Source

Dependent
Variable

Type III
Sum of

df

Squares

Mean
Square

Sig.

Corrected Model PERC_SEV

.667a

.667

.782

.382

PERC_VUL

.711b

.711

.343

.561

PERC_SEV

1715.810

1715.810

2012.613 .000

PERC_VUL

940.055

940.055

453.602

.000

TIME_FACTOR PERC_SEV

.667

.667

.782

.382

PERC_VUL

.711

.711

.343

.561

PERC_SEV

34.101

40

.853

PERC_VUL

82.897

40

2.072

PERC_SEV

1751.250

42

PERC_VUL

1023.333

42

PERC_SEV

34.768

41

PERC_VUL

83.608

41

Intercept

Error

Total

Corrected Total

a. R Squared = .019 (Adjusted R Squared = -.005)


b. R Squared = .009 (Adjusted R Squared = -.016)

Hypothesis 2 was not supported as geographical proximity had no


effect on perceived severity and vulnerability (Pillais Trace=0.118; Wilks
lambda = 0.882; Hotellings Trace and Roys Largest Root = 0.134,
F(2,39)=2.60). The results are shown in Tables 6.18a and 6.18b.

132

Table 6.18a Experiment 1: Hypothesis 2: multivariate tests (watch


stimuli)
Multivariate Testsb
Effect
Intercept

Value

Pillai's Trace
Wilks' Lambda
Hotelling's Trace

Hypothesis df Error df Sig.

.981 990.045

2.000

39.000 .000

.019 990.045

2.000

39.000 .000

50.772 990.045a

2.000

39.000 .000

2.000

39.000 .000

Roy's Largest Root 50.772 990.045


GEOG_FACTOR Pillai's Trace
Wilks' Lambda
Hotelling's Trace

.118

2.607a

2.000

39.000 .087

.882

2.607

2.000

39.000 .087

2.607

2.000

39.000 .087

2.607

2.000

39.000 .087

.134

Roy's Largest Root .134

a. Exact statistic b. Design: Intercept + GEOG_FACTOR

Table 6.18b Experiment 1: Hypothesis 2: tests of between-subjects effects


(watch stimuli)
Tests of Between-Subjects Effects
Source
Corrected Model

Dependent
Variable
PERC_SEV

Type III
Sum of
Squares

df

Mean
Square

Sig.

2.650a

2.650

3.301

.077

.502

.242

.626

PERC_VUL

.502

PERC_SEV

1629.079

1629.079

2028.889 .000

PERC_VUL

913.772

913.772

439.808

.000

GEOG_FACTOR PERC_SEV

2.650

2.650

3.301

.077

PERC_VUL

.502

.502

.242

.626

PERC_SEV

32.118

40

.803

PERC_VUL

83.106

40

2.078

PERC_SEV

1751.250

42

PERC_VUL

1023.333

42

PERC_SEV

34.768

41

PERC_VUL

83.608

41

Intercept

Error
Total
Corrected Total

a. R Squared = .076 (Adjusted R Squared = .053)


b. R Squared = .006 (Adjusted R Squared = -.019)

133

A 2 (temporal proximity of threat: day vs.year) x 2 (geographical


proximity of threat: India vs. World) multivariate analysis of variance
(MANOVA) was conducted to examine the interaction effects between
temporal proximity and geographical proximity to verify H3. It can be seen
from Table 6.19a and Table 6.19b that no interaction effect exists between
temporal proximity of threat and geographical proximity of threat on
perceived severity and vulnerability (Pillais Trace=0.044; Wilks lambda =
0.956; Hotellings Trace and Roys Largest Root = 0.046, F(2,37)=0.858).
Therefore, H3 was not accepted and the results are shown in
Table 6.19a Experiment 1: Hypothesis 3: multivariate tests (watch
stimuli)
Multivariate Tests b
Effect
Intercept

2.000

37.000 .000

.018 1003.743

2.000

37.000 .000

54.256 1003.743

2.000

37.000 .000

Roy's Largest Root 54.256 1003.743a

Hotelling's Trace
Pillai's Trace
Wilks' Lambda
Hotelling's Trace

.129
.871
.148

Roy's Largest Root .148


TIME_FACTOR

Pillai's Trace
Wilks' Lambda
Hotelling's Trace

.044
.956
.046

Roy's Largest Root .046


GEOG_FACTOR *
TIME_FACTOR

Hypothesis Error
Sig.
df
df

.982 1003.743a

Pillai's Trace
Wilks' Lambda

GEOG_FACTOR

Value

2.000

37.000 .000

2.735

2.000

37.000 .078

2.735

2.000

37.000 .078

2.735

2.000

37.000 .078

2.735

2.000

37.000 .078

.858

2.000

37.000 .432

.858

2.000

37.000 .432

.858

2.000

37.000 .432

.858

2.000

37.000 .432

2.000

37.000 .372

Pillai's Trace

.052

1.017

Wilks' Lambda

.948

1.017a

2.000

37.000 .372

.055

1.017

2.000

37.000 .372

1.017

2.000

37.000 .372

Hotelling's Trace

Roy's Largest Root .055

a. Exact statistic
b. Design: Intercept + GEOG_FACTOR + TIME_FACTOR + GEOG_FACTOR *
TIME_FACTOR

134

Table 6.19b Experiment 1: Hypothesis 3: tests of between-subjects effects


(watch stimuli)
Tests of Between-Subjects Effects
Dependent
Variable

Type III
Sum of
Squares

df

Mean
Square

Sig.

PERC_SEV

5.079a

1.693

2.167

.108

PERC_VUL

1.233b

.411

.190

.903

PERC_SEV

1610.267

1610.267

2061.013 .000

PERC_VUL

906.243

906.243

418.050

.000

PERC_SEV

2.853

2.853

3.651

.064

PERC_VUL

.412

.412

.190

.665

PERC_SEV

1.377

1.377

1.762

.192

PERC_VUL

.695

.695

.321

.575

GEOG_FACTOR *
TIME_FACTOR

PERC_SEV

1.500

1.500

1.920

.174

PERC_VUL

.116

.116

.053

.819

Error

PERC_SEV

29.689

38

.781

PERC_VUL

82.376

38

2.168

PERC_SEV

1751.250

42

PERC_VUL

1023.333

42

PERC_SEV

34.768

41

PERC_VUL

83.608

41

Source
Corrected Model
Intercept
GEOG_FACTOR
TIME_FACTOR

Total
Corrected Total

a. R Squared = .146 (Adjusted R Squared = .079)


b. R Squared = .015 (Adjusted R Squared = -.063)

To test H4 a one-way MANCOVA was conducted with perceived


severity and perceived vulnerability as dependent variables and CFC as the
covariate. It can be inferred from Tables 6.20a and 6.20b that there was no
statistically significant difference between the groups and hence H4 was also
not supported.

135

Table 6.20a Experiment 1: Hypothesis 4: multivariate tests (watch


stimuli)
Multivariate Testsb
Effect
Value F Hypothesis df Error df Sig.
Intercept
Pillai's Trace
.587 26.961a
2.000
38.000 .000
a
Wilks' Lambda
.413 26.961
2.000
38.000 .000
a
Hotelling's Trace 1.419 26.961
2.000
38.000 .000
a
Roy's Largest Root 1.419 26.961
2.000
38.000 .000
CFC_TOTAL Pillai's Trace
.086 1.791a
2.000
38.000 .181
a
Wilks' Lambda
.914 1.791
2.000
38.000 .181
a
Hotelling's Trace .094 1.791
2.000
38.000 .181
a
Roy's Largest Root .094 1.791
2.000
38.000 .181
a
TIME_FACTOR Pillai's Trace
.024 .467
2.000
38.000 .630
Wilks' Lambda
.976 .467a
2.000
38.000 .630
a
Hotelling's Trace .025 .467
2.000
38.000 .630
a
Roy's Largest Root .025 .467
2.000
38.000 .630
a. Exact statistic
b. Design: Intercept + CFC_TOTAL + TIME_FACTOR

Table 6.20b Experiment 1: Hypothesis 4: tests of between-subjects effects


(watch stimuli)
Tests of Between-Subjects Effects
Dependent
Type III Sum
Mean
Source
df
Variable
of Squares
Square
Corrected Model PERC_SEV
1.441a
2 .720
b
PERC_VUL
7.717
2 3.859
Intercept
PERC_SEV
47.057
1 47.057
PERC_VUL
9.716
1 9.716
CFC_TOTAL
PERC_SEV
.774
1 .774
PERC_VUL
7.006
1 7.006
TIME_FACTOR PERC_SEV
.712
1 .712
PERC_VUL
.859
1 .859
Error
PERC_SEV
33.327
39 .855
PERC_VUL
75.891
39 1.946
Total
PERC_SEV
1751.250
42
PERC_VUL
1023.333
42
Corrected Total
PERC_SEV
34.768
41
PERC_VUL
83.608
41
a. R Squared = .041 (Adjusted R Squared = -.008)
b. R Squared = .092 (Adjusted R Squared = .046)

Sig.

.843
1.983
55.068
4.993
.906
3.600
.834
.441

.438
.151
.000
.031
.347
.065
.367
.510

136

Similar to the previous experiment, a three way ANOVA was


conducted to analyse the effect of the independent variables and covariate on
perceived severity and perceived vulnerability. The results are shown in
Tables 6.21a and 6.21b. Unlike the mobile phone stimuli, the factors and the
covariate did not influence perceived severity or perceived vulnerability.
Table 6.21a Experiment 1: Tests of between-subjects effects (effect of
temporal, geographical proximity of threat and cfc on
perceived severity)
Tests of Between-Subjects Effects
Dependent Variable:PERC_SEV
Source

Type III
Mean
Sum of df
Square
Squares

Partial
Noncent. Observed
Sig. Eta
Parameter Powerb
Squared

Corrected Model

6.298a

6 1.050 1.291 .287

.181

7.743

.437

Intercept

15.534 1 15.534 19.097 .000

.353

19.097

.989

TIME_FACTOR

.084

.084

.103 .750

.003

.103

.061

GEOG_FACTOR

.608

.608

.747 .393

.021

.747

.134

CFC_TOTAL

1.295

1 1.295 1.591 .215

.043

1.591

.233

TIME_FACTOR
* CFC_TOTAL

.266

.266

.328 .571

.009

.328

.086

GEOG_FACTOR
* CFC_TOTAL

.321

.321

.395 .534

.011

.395

.094

TIME_FACTOR*
GEOG_FACTOR
* CFC_TOTAL

1.520

1 1.520 1.869 .180

.051

1.869

.265

Error

28.470 35 .813

Total
Corrected Total

1751.250 42
34.768 41

a. R Squared = .181 (Adjusted R Squared = .041)


b. Computed using alpha = .05

137

Table 6.21b Experiment 1: Tests of between-subjects effects (effect of


temporal, geographical proximity of threat and cfc on
perceived vulnerability)
Tests of Between-Subjects Effects
Dependent Variable: PERC_VUL
Source

Type III
Mean
Sum of df
Square
Squares

Partial
Noncent. Observed
Sig.
Eta
Parameter Powerb
Squared

11.303a

1.884

.912 .498

.135

5.471

.310

Intercept

5.637

5.637 2.729 .108

.072

2.729

.362

TIME_FACTOR

3.258

3.258 1.577 .218

.043

1.577

.231

GEOG_FACTOR

.017

.017

.008 .928

.000

.008

.051

CFC_TOTAL

2.046

2.046

.990 .327

.028

.990

.162

2.738

2.738 1.326 .257

.036

1.326

.201

GEOG_FACTOR
* CFC_TOTAL

.040

.040

.019 .890

.001

.019

.052

TIME_FACTOR *
GEOG_FACTOR
* CFC_TOTAL

.304

.304

.147 .703

.004

.147

.066

Corrected Model

TIME_FACTOR *
CFC_TOTAL

Error

72.305 35 2.066

Total

1023.333 42

Corrected Total

83.608 41

a. R Squared = .135 (Adjusted R Squared = -.013)


b. Computed using alpha = .05

The following table (Table 6.22) summarizes the effects of the


manipulations on the PMT variables.

138

Table 6.22

Experiment 1: Summary of the hypotheses (H1 H4) -watch


stimulus

Hypothesis

Factor

Perceived
severity

Perceived
vulnerability

H1

Temporal proximity

H2

Geographical proximity

H3

Temporal proximity *
Geographical proximity

H4

CFC

X no effect

6.3.6.4

- Effect present

Hypotheses tests of the relationship among PMT variables,


involvement, attitudes and intentions
Standard simple or multiple linear regression (ordinary least

squares (OLS)) analysis was run to ascertain the relationship between the
PMT variables, involvement, attitudes and intention variables. Only one or
two predictors were considered at a time. Similarly assumptions regarding
multicollinearity, homoscedascity, linearity and normality of residuals were
met in all the scenarios
Mobile phone stimuli
A multiple regression analysis was conducted to evaluate how well
the perceived severity and perceived vulnerability predicted fear and the
efficacy variables. Neither perceived severity nor perceived vulnerability
predicted the dependent variables. Therefore H9a, H9b and H9c were not
supported. The results are shown in Table 6.23a, 6.23b and 6.23c. It can
therefore be inferred that the perceived threat levels did not affect fear.

139

Table 6.23a Experiment 1: Hypothesis 9a: effect of threat appraisal


components on fear (mobile phone stimuli)
Model Summary
Model
1

R
.280

R Square Adjusted R Square Std. Error of the Estimate


a

.079

.046

.90543

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

3.919

1.960

2.390

.101a

Residual

45.909

56

.820

Total

49.828

58

Model
1

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: FEAR

Table 6.23b Experiment 1: Hypothesis 9b: effect of threat appraisal


components on response efficacy (mobile phone stimuli)
Model Summary
Model
1

R
.247

R Square Adjusted R Square Std. Error of the Estimate


a

.061

.027

1.23681

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression 5.543

2.771

Residual

85.663

56

1.530

Total

91.205

58

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: RESP_EFFICACY

F
1.812

Sig.
.173a

140

Table 6.23c Experiment 1: Hypothesis 9c: effect of threat appraisal


components on self efficacy (mobile phone stimuli)
Model Summary
Model

.074a

R Square Adjusted R Square Std. Error of the Estimate


0.005

-0.030

1.16446

a. Predictors: (Constant), PERC_VUL, PERC_SEV

ANOVAb
Model
1

Regression

Sum of
Squares

df

Mean
Square

.419

.210

Residual

75.935

56

1.356

Total

76.354

58

Sig.

.155

.857a

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: SELF_EFFICACY

Simple linear regression analyses was conducted to evaluate


how well fear and the other PMT variables (perceived severity, vulnerability,
response efficacy and self efficacy) predict message involvement. The results
are shown in the following tables (Tables 6.24a, 6.24b, 6.24c, 6.24d and
6.24e). It can be seen from Table 6.24b that perceived vulnerability predicts
message involvement ( =0.30 t(58)=2.43, p<0.05, R2=0.09). From Table
6.24d, it can be seen that response efficacy predicted message involvement
=0.30 t(58)=2.41, p<0.05, R2=0.09). Therefore only H10b and H10d were
supported. The other PMT variables did not have any effect on message
involvement. Hence only perceived vulnerability and response efficacy have
an effect on message involvement.

141

Table 6.24a Experiment 1: Hypothesis 10a: effect of perceived severity


on message involvement (mobile phone stimuli)
Model Summary
Model

R Square Adjusted R Square Std. Error of the Estimate

.197

0.039

0.022

1.06258

a. Predictors: (Constant), PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

2.596

2.596

2.299

.135a

Residual

64.357

57

1.129

Total

66.953

58

Model
1

a. Predictors: (Constant), PERC_SEV


b. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.24b Experiment 1: Hypothesis 10b: effect of perceived


vulnerability on message involvement (mobile phone
stimuli)
Model Summary
Model

R Square

Adjusted R
Square

Std. Error of the


Estimate

.307a

0.094

0.078

1.03138

a. Predictors: (Constant), PERC_VUL


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

6.319

6.319

5.941

.018a

Residual

60.634

57

1.064

Total

66.953

58

Model
1

a. Predictors: (Constant), PERC_VUL


b. Dependent Variable: MESSAGE_INVOLVEMENT

142

Table 6.24b (Continued)


Coefficientsa
Unstandardized

Standardized

Coefficients

Coefficients

Model

1 (Constant)
PERC_VUL

Std. Error

3.602

.534

.269

.110

Sig.

Beta
6.743 .000
.307

2.437 .018

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.24c Experiment 1: Hypothesis 10c: effect of fear on message


involvement (mobile phone stimuli)
Model Summary
Model

.193a

R Square Adjusted R Square


0.037

Std. Error of the


Estimate

0.020

1.06344

a. Predictors: (Constant), FEAR


ANOVAb
Model
1

Sum of

df

Squares

Mean
Square

Regression 2.491

2.491

Residual

64.461

57

1.131

Total

66.953

58

a. Predictors: (Constant), FEAR


b. Dependent Variable: MESSAGE_INVOLVEMENT

F
2.203

Sig.
.143a

143

Table 6.24d Experiment 1: Hypothesis 10d: effect of response efficacy on


message involvement (mobile phone stimuli)
Model Summary
Model

R Square Adjusted R Square Std. Error of the Estimate

.305a

0.093

0.077

1.032228

a. Predictors: (Constant), RESP_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

6.214

6.214

5.831

.019a

Residual

60.739

57

1.066

Total

66.953

58

Model
1

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: MESSAGE_INVOLVEMENT

Coefficientsa
Unstandardized
Coefficients

Model

1 (Constant)
RESP_EFFICACY

Std. Error

3.561

.555

.261

.108

Standardized
Coefficients

Sig.

Beta
6.415 .000
.305

2.415 .019

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.24e Experiment 1: Hypothesis 10e: effect of self efficacy on


message involvement (mobile phone stimuli)
Model Summary
Model
1

R
.167

R Square Adjusted R Square Std. Error of the Estimate


a

0.028

0.011

a. Predictors: (Constant), SELF_EFFICACY

1.06864

144

Table 6.24e (Continued)


ANOVAb
Model
1

Regression
Residual
Total

Sum of
Squares
1.859
65.094
66.953

df
1
57
58

Mean
Square
1.859

Sig.

1.628

.207a

1.142

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: MESSAGE_INVOLVEMENT

Simple regression analysis was conducted to evaluate how well


coping variables predicted attitude towards the ad and purchase intention. It
can be seen from Table 6.25a and Table 6.25c that response efficacy
significantly predicted attitude towards the ad ( =0.32 t(58)=2.550, p<0.05)
and purchase intention ( =0.297 t(58)=2.352, p<0.05 R2=0.08) while selfefficacy did not predict both the variables (Tables 6.25b and Table 6.25d).
Therefore H11a and H11c were supported and H11b and H11d were not
supported.
Table 6.25a Experiment 1: Hypothesis 11a: effect of response efficacy on
attitude towards ad (mobile phone stimuli)
Model Summary
Model
1

R
.320

R Square Adjusted R Square Std. Error of the Estimate


a

0.102

0.087

1.04128

a. Predictors: (Constant), RESP_EFFICACY


ANOVAb
Model
1

Regression
Residual
Total

Sum of
Squares
7.048
61.803
68.851

df
1
57
58

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: ATTITUDE_AD

Mean
Square
7.048
1.084

Sig.

6.500

.013a

145

Table 6.25a (Continued)


Coefficientsa

Model

Unstandardized

Standardized

Coefficients

Coefficients

Std. Error

(Constant)

3.858

.560

RESP_EFFICACY

.278

.109

Sig.

Beta
6.889 .000
.320

2.550 .013

a. Dependent Variable: ATTITUDE_AD

Table 6.25b Experiment 1: Hypothesis 11b: effect of self efficacy on


attitude towards ad (mobile phone stimuli)
Model Summary
Model

.126a

R Square Adjusted R Square Std. Error of the Estimate


0.016

-0.001

1.09033

a. Predictors: (Constant), SELF_EFFICACY

ANOVAb
Model
1

Regression

Sum of
Squares

Mean
Square

df

1.089

1.089

Residual

67.762

57

1.189

Total

68.851

58

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: ATTITUDE_AD

F
.916

Sig.
.343a

146

Table 6.25c Experiment 1: Hypothesis 11c: effect of response efficacy on


purchase intention (mobile phone stimuli)
Model Summary
Model
1

R
.297

R Square Adjusted R Square Std. Error of the Estimate


a

0.088

0.072

1.66704

a. Predictors: (Constant), RESP_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

15.373

15.373

5.532

.022a

Residual

158.405

57

2.779

Total

173.778

58

Model
1

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: PURCHASE_INTENTION
Coefficientsa
Unstandardized
Coefficients

Model

B
1

(Constant)

1.954

RESP_EFFICACY .411

Standardized
Coefficients

Std.
Error

Sig.

Beta

.896
.175

.297

2.180

.033

2.352

.022

a. Dependent Variable: PURCHASE_INTENTION

Table 6.25d Hypothesis 11d: effect of self efficacy on purchase intention


(mobile phone stimuli)
Model Summary
Model

.0.215a

R Square Adjusted R Square


0.46

0.030

a. Predictors: (Constant), SELF_EFFICACY

Std. Error of the


Estimate
1.70519

147

Table 6.25d (Continued)


ANOVAb

Regression

Sum of
Squares
8.041

Residual
Total

165.737
173.778

Model
1

Mean
Square
8.041

df
1
57
58

Sig.

2.765

.102a

2.908

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: PURCHASE_INTENTION

H12a was not supported as environmental knowledge predicted


perceived severity in the direction opposite to the one hypothesized. H12b and
H12c were not supported as simple regression analyses revealed that
environmental knowledge did not predict perceived vulnerability and fear.
Environmental knowledge did not predict message involvement and therefore
H12d was not supported. Tables 6.26a, 6.26b 6.26c, 6.26d illustrate the results.
Table 6.26a Experiment 1: Hypothesis 12a: effect of environmental
knowledge on perceived severity (mobile phone stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.333a

0.111

0.095

0.70162

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression

3.496

3.496

Residual

28.059

57

.492

Total

31.555

58

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: PERC_SEV

F
7.102

Sig.
.010a

148

Table 6.26a (Continued)


Coefficientsa
Unstandardized
Coefficients

Model

Std.
Error

(Constant)

5.053

.338

ENV_KNOW

.108

.041

Standardized
Coefficients

Sig.

14.954

.000

2.665

.010

Beta

.333

a. Dependent Variable: PERC_SEV

Table 6.26b Experiment 1: Hypothesis 12b: effect of environmental


knowledge on perceived vulnerability (mobile phone stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.033a

0.001

-0.016

1.23849

a. Predictors: (Constant), ENV_KNOW

ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression .096

.096

Residual

87.429

57

1.534

Total

87.525

58

F
.063

Sig.
.803a

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: PERC_VUL

The results show that environmental knowledge played no role in


influencing threat perception.

149

Table 6.26c Experiment 1: Hypothesis 12c: effect of environmental


knowledge on fear (mobile phone stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.045a

0.002

-0.015

0.93401

a. Predictors: (Constant), ENV_KNOW

ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.103

.103

.118

.733a

Residual

49.725

57

.872

Total

49.828

58

Model
1

Regression

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: FEAR

Table 6.26d Experiment 1: Hypothesis 12d: effect of environmental


knowledge on message involvement (mobile phone stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.065a

0.004

-0.013

1.08150

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Model
1

Regression
Residual
Total

Sum of
Squares

df

Mean
Square

Sig.

.283

.283

.242

.625a

66.670

57

1.170

66.953

58

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: MESSAGE_INVOLVEMENT

150

Similar regression analysis was conducted to evaluate the


relationship among involvement variables and attitude towards ad.
Environmental concern was positively related to message involvement and
therefore H13a was supported ( =0.267, t(58)=2.055, p<0.05 R2=0.07).
However it did not predict attitude towards the ad and purchase intention.
Therefore H13b and H13c were not supported. The following tables (Tables
6.27a, 6.27b, 6.27c) show the regression results which clearly show that
environmental concern was related only to message involvement.
Table 6.27a Experiment 1: Hypothesis 13a: effect of environmental
concern (enduring involvement with the environment) on
message involvement (mobile phone stimuli)
Model Summary
Model
1

R
0.267

R Square Adjusted R Square Std. Error of the Estimate


a

0.071

0.054

1.01307

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Model
1

Regression
Residual
Total

Sum of
Squares
4.336
56.447
60.784

Mean
Square
4.336
1.026

df
1
55
56

Sig.

4.225

.045a

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa

Model

Unstandardized Standardized
Coefficients
Coefficients
B

1 (Constant)

3.021

TOTAL_ENV_CONCERN .324

Std.
Error

Sig.

Beta

.934
.158

3.234 .002
.267

a. Dependent Variable: MESSAGE_INVOLVEMENT

2.055 .045

151

Table 6.27b Experiment 1: Hypothesis 13b: effect of environmental


concern (enduring involvement with the environment) on
attitude towards the ad (mobile phone stimuli)
Model Summary
Model

R Square Adjusted R Square Std. Error of the Estimate

.0.119a

0.014

-0.004

1.09816

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.947

.947

.785

.379a

Residual

66.328

55

1.206

Total

67.275

56

Model
1

Regression

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: ATTITUDE_AD

Table 6.27c Experiment 1: Hypothesis 13c: effect of environmental


concern (enduring involvement with the environment) on
purchase intention (mobile phone stimuli)
Model Summary
Model
1

R
0.012

R Square Adjusted R Square Std. Error of the Estimate


a

0.000

-0.018

1.70730

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.022

.022

.007

.932a

Residual

160.318

55

2.915

Total

160.339

56

Model
1

Regression

a. Predictors: (Constant), TOTAL_ENV_CONCERN b. Dependent Variable:


PURCHASE_INTENTION

152

H14 was supported as message involvement predicted attitude


towards ad ( =0.616, t(58)=5.902, p<0.001). Overall model fit was R2=0.359.
Table 6.28 shows the results.
Table 6.28

Experiment 1: Hypothesis 14: effect of message involvement


on attitude towards ad (mobile phone stimuli)
Model Summary

Model R
1

R Square Adjusted R Square Std. Error of the Estimate

0.616 a

0.359

0.368

0.86585

a. Predictors: (Constant), MESSAGE_INVOLVEMENT


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

26.118

26.118

34.838

.000a

Residual

42.733

57

.750

Total

68.851

58

Model
1

a. Predictors: (Constant), MESSAGE_INVOLVEMENT


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Model
1 (Constant)
MESSAGE_INVOLVEMENT

Unstandardized
Coefficients
B

Std. Error

2.207

.527

.625

.106

Standardized
Coefficients

Sig.

Beta
4.190 .000
.616

5.902 .000

a. Dependent Variable: ATTITUDE_AD

It can be seen from Tables 6.29 and 6.30 that H15 and H16 were
also supported as attitude towards the ad significantly predicted the attitude
towards the brand ( =0.58 t(58)=5.412, p<0.001, R2=0.33) and attitude

153

towards the brand significantly predicted the purchase intention ( =0.517,


t(58)=4.557, p<0.001, R2=0.267).
Table 6.29

Experiment 1: Hypothesis 15: effect of attitude towards ad


on attitude towards brand (mobile phone stimuli)
Model Summary

Model

0.583 a

R Square Adjusted R Square Std. Error of the Estimate


0.339

0.328

0.91829

a. Predictors: (Constant), ATTITUDE_AD


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

24.703

24.703

29.295

.000a

Residual

48.065

57

.843

Total

72.768

58

Model
1

a. Predictors: (Constant), ATTITUDE_AD


b. Dependent Variable: ATTITUDE_BRAND
Coefficientsa

Model

Unstandardized
Coefficients
B

Std. Error

(Constant)

1.905

.592

ATTITUDE_AD

.599

.111

a. Dependent Variable: ATTITUDE_BRAND

Standardized
Coefficients

Sig.

3.215

.002

5.412

.000

Beta

.583

154

Table 6.30

Experiment 1: Hypothesis 16: effect of attitude towards


brand on purchase intention (mobile phone stimuli)
Model Summary

Model

0.517

R Square Adjusted R Square Std. Error of the Estimate


0.267

0.254

1.49485

a. Predictors: (Constant), ATTITUDE_BRAND


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

46.406

46.406

20.767

.000a

Residuals

127.372

57

2.235

Total

173.778

58

Model
1

a. Predictors: (Constant), ATTITUDE_BRAND


b. Dependent Variable: PURCHASE_INTENTION
Coefficientsa
Model

Unstandardized
Coefficients
B

Std. Error

(Constant)

-.029

.905

ATTITUDE_BRAND

.799

.175

Standardized
Coefficients

Sig.

-.032

.975

4.557

.000

Beta

.517

a. Dependent Variable: PURCHASE_INTENTION

The results of the hypothesis tests regarding the relationship


between PMT variables, involvement, attitudes and intentions for the mobile
phone stimuli are summarised in Table 6.31.

155

Table 6.31

Hypothesis
H9a

Experiment 1: Summary of hypotheses tests regarding


relationship between PMT variables, involvement, attitudes
and intentions (H9 H16) (mobile phone stimuli)
Predictor

Perceived
Severity and

Dependent
variable

R2

Adjusted Unstandardised Standardised


R2
coefficient B coefficient

Fear

Not
Significant

Response
Efficacy

Not
Significant

Self Efficacy

Not
Significant

Not
Significant

Perceived
vulnerability
H9b

Perceived
Severity and
Perceived
vulnerability

H9c

Perceived
Severity and
Perceived
vulnerability

H10a

Perceived
Severity

Message
Involvement

H10b

Perceived

Message
0.09 0.07
Involvement

Vulnerability
H10c

Fear

Message
Involvement

H10d

Response
Efficacy

Message
0.09 0.07
Involvement

H10e

Self Efficacy

Message
Involvement

H11a

Response
efficacy

Attitude
towards ad

H11b

Self Efficacy

Attitude
towards ad

H11c

Response
efficacy

Purchase
Intention

H11d

Self Efficacy

Purchase
Intention

H12a

Environmental Perceived
Severity
Knowledge

0.26

0.30*
Not
Significant

0.26

0.30*
Not
Significant

0.10 0.08

0.27

0.32*
Not
Significant

0.08 0.07

0.41

0.29*
Not
Significant

0.11 0.09

0.10

0.33*

156

Table 6.31 (Continued)

H12b

Environmental Perceived
Vulnerability
Knowledge

Not
Significant

H12c

Environmental Fear
Knowledge

Not
Significant

H12d

Environmental Message
Involvement
Knowledge

Not
Significant

H13a

Environmental Message
0.07 0.05
concern
Involvement

H13b

Environmental Attitude
concern
towards ad

Not
significant

H13c

Environmental Purchase
concern
Intention

Not
significant

H14

Message
Involvement

Attitude
towards ad

0.37 0.36

0.625

0.616***

H15

Attitude
towards ad

Attitude
towards
brand

0.33 0.32

0.59

0.58***

H16

Attitude
Purchase
towards brand Intention

0.26 0.25

0.79

0.51***

0.32

0.26*

***p <.001 **p <.01 *p <.05; n= 58

Watch stimuli
A multiple regression analysis was conducted to evaluate how well
perceived severity and perceived vulnerability predicted fear and response
efficacy. Table 6.32a shows that H9a was not supported as the model was not
significant. H9b was partially supported as perceived vulnerability positively
influenced response efficacy ( =0.595 t(39)=4.099, p<0.001, R2 =0.309). H9c
was not tested because of the low reliability values for self efficacy. The
results of the regression analysis are shown below in Table 6.32b.

157

Table 6.32a Experiment 1: Hypothesis 9a: effect of threat appraisal


components on fear (watch stimuli)
Model Summary
Model
1

R
.153

R Square Adjusted R Square Std. Error of the Estimate


a

0.023

-0.27

1.00674

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.949

.474

.468

.630a

Residual

39.527

39

1.014

Total

40.476

41

Model
1

Regression

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: FEAR

Table 6.32b Experiment 1: Hypothesis 9b: effect of threat appraisal


components on response efficacy (watch stimuli)
Model Summary
Model R
1

.556

R Square Adjusted R Square Std. Error of the Estimate


a

0.309

0.273

1.05815

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

19.489

9.744

8.703

.001a

Residual

43.667

39

1.120

Total

63.156

41

Model
1

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: RESP_EFFICACY

158

Table 6.32b (Continued)


Coefficientsa
Unstandardized
Coefficients

Model
1

Standardized
Coefficients

Sig.

3.313

.002

Std. Error

Beta

(Constant)

3.854

1.163

PERC_SEV

-.180

.196

-.133

-.919

.364

PERC_VUL

.517

.126

.595

4.099

.000

a. Dependent Variable: RESP_EFFICACY

H10a was supported as perceived severity predicted message


involvement ( =0.37 t(39)=2.496, p<0.05, R2 = 0.367). H10b and H10d were
supported, as simple linear regression analyses revealed that among the PMT
variables only perceived vulnerability ( =0.54 t(39)=4.108, p<0.001, R2 =
0.29) and response efficacy ( =0.39 t(39)=2.717, p<0.05,R2 = 0.156)
significantly predicted message involvement. The results are shown in the
following tables (Tables 6.33a, 6.33b, 6.33c and 6.33d). H10e was not tested
because it involved self efficacy.
Table 6.33a Experiment 1: Hypothesis 10a: effect of perceived severity
on message involvement (watch stimuli)
Model Summary
Model
1

R
.367

R Square Adjusted R Square Std. Error of the Estimate


a

0.135

0.113

1.02271

a. Predictors: (Constant), PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

6.518

6.518

6.232

.017a

Residual

41.837

40

1.046

Model
1

159

Table 6.33a (Continued)


Model
Total

Sum of
Squares

df

48.355

41

Mean
Square

Sig.

a. Predictors: (Constant), PERC_SEV


b. Dependent Variable: MESSAGE_INVOLVEMENT

Coefficientsa
Unstandardized
Coefficients

Model

B
1

(Constant)

Standardized
Coefficients

Std. Error

2.077

.173

Sig.

Beta

1.120

PERC_SEV .433

.367

1.855

.071

2.496

.017

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.33b Experiment 1: Hypothesis 10b: effect of perceived


vulnerability on message involvement (watch stimuli)
Model Summary
Model R
1

.545

R Square Adjusted R Square Std. Error of the Estimate


a

0.297

0.279

0.92206

a. Predictors: (Constant), PERC_VUL


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

14.348

14.348

16.876

.000a

Residual

34.008

40

.850

Total

48.355

41

Model
1

a. Predictors: (Constant), PERC_VUL


b. Dependent Variable: MESSAGE_INVOLVEMENT

160

Table 6.33b (Continued)


Coefficientsa
Unstandardized
Coefficients

Model

B
1

(Constant)

Standardized
Coefficients

Std. Error

2.886

.101

Sig.

Beta

.498

PERC_VUL .414

.545

5.798

.000

4.108

.000

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.33c Experiment 1: Hypothesis 10c: effect of fear on message


involvement (watch stimuli)
Model Summary
Model R
1

R Square Adjusted R Square Std. Error of the Estimate

.248a

0.061

0.038

1.06524

a. Predictors: (Constant), FEAR


ANOVAb
Sum of
Squares

Model
1

Mean
Square

df

Regression 2.966

2.966

Residual

45.389

40

1.135

Total

48.355

41

F
2.614

Sig.
.114a

a. Predictors: (Constant), FEAR


b. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.33d Experiment 1: Hypothesis 10d: effect of response efficacy on


message involvement (watch stimuli)
Model Summary
Model R
1

.395

R Square Adjusted R Square Std. Error of the Estimate


a

0.156

0.135

a. Predictors: (Constant), RESP_EFFICACY

1.01021

161

Table 6.33d (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

7.534

7.534

7.383

.010a

Residual

40.821

40

1.021

Total

48.355

41

Model
1

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: MESSAGE_INVOLVEMENT

Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

3.066

.673

RESP_EFFICACY

.345

.127

Standardized
Coefficients

Sig.

4.556

.000

2.717

.010

Beta
.395

a. Dependent Variable: MESSAGE_INVOLVEMENT

Results of the simple regression analysis revealed the effect of


response efficacy on attitude towards the ad and purchase intention. The
model was not significant. Table 6.34a and 6.34b show that H11a and H11c
were not supported as response efficacy did not predict attitude towards the ad
and purchase intention.
Table 6.34a Experiment 1: Hypothesis 11a: effect of response efficacy on
attitude towards ad (watch stimuli)
Model Summary
Model R
1

.200a

R Square Adjusted R Square Std. Error of the Estimate


0.040

0.016

a. Predictors: (Constant), RESP_EFFICACY

1.01291

162

Table 6.34a (Continued)

ANOVAb
Sum of
Squares

Model
1

Mean
Square

df

Regression 1.714

1.714

Residual

41.040

40

1.026

Total

42.754

41

F
1.671

Sig.
.204a

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: ATTITUDE_AD

Table 6.34b Experiment 1: Hypothesis 11c: effect of response efficacy on


purchase intention (watch stimuli)
Model Summary
Model R
1

.266

R Square Adjusted R Square Std. Error of the Estimate


a

0.071

0.047

1.50427

a. Predictors: (Constant), RESP_EFFICACY

ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression 6.865

6.865

Residual

90.513

40

2.263

Total

97.378

41

F
3.034

Sig.
.089a

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: PURCHASE_INTENTION

H12a, H12b, H12c and H12d were not supported environmental


knowledge did not predict the hypothesized dependent variables. Tables 6.35a,
6.35b 6.35c, 6.35d illustrate the results.

163

Table 6.35a Experiment 1: Hypothesis 12a: effect of environmental


knowledge on perceived severity (watch stimuli)
Model Summary
Model R
R Square Adjusted R Square Std. Error of the Estimate
a
1
.0.247
0.061
0.038
0.90330
a. Predictors: (Constant), ENV_KNOW
ANOVAb
Model
1

Regression
Residual

Sum of
Squares
2.130
32.638

Mean
Square
2.130

40

.816

df

Sig.

2.610

.114a

Total
34.768
41
a. Predictors: (Constant), ENV_KNOW
b. Dependent Variable: PERC_SEV

Table 6.35b Experiment 1: Hypothesis 12a: effect of environmental


knowledge on perceived vulnerability (watch stimuli)
Model Summary
Model R
R Square Adjusted R Square Std. Error of the Estimate
a
1
.0.219 0.048
0.024
1.41063
a. Predictors: (Constant), ENV_KNOW
ANOVAb
Model
1

Regression
Residual
Total

Sum of
Squares
4.014
79.595
83.608

df
1
40
41

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: PERC_VUL

Mean
Square
4.014
1.990

Sig.

2.017

.163a

164

Table 6.35c Experiment 1: Hypothesis 12a: effect of environmental


knowledge on fear (watch stimuli)
Model Summary
Model
1

R
0.123

R Square Adjusted R Square Std. Error of the Estimate


a

0.015

-0.10

0.99834

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.609

.609

.611

.439a

Residual

39.867

40

.997

Total

40.476

41

Model
1

Regression

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: FEAR

Table 6.35d Experiment 1: Hypothesis 12b: effect of environmental


knowledge on message involvement (watch stimuli)
Model Summary
Model
1

R Square Adjusted R Square Std. Error of the Estimate


a

.0.010

0.000

-0.025

1.09944

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression .005

.005

Residual

48.350

40

1.209

Total

48.355

41

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: MESSAGE_INVOLVEMENT

F
.004

Sig.
.950a

165

Simple regression analysis was conducted to evaluate if


involvement predicted attitude towards ad. It can be seen from Tables 6.36a,
6.36b, 6.36c that the regression model was not significant. Unlike the
previous experiment with mobile stimuli, environmental concern was not
related to message involvement, attitude towards ad or purchases intentions.
Therefore H13a, H13b and H13c were not supported.
Table 6.36a Experiment 1: Hypothesis 13a: effect of environmental
concern (enduring involvement with the environment) on
message involvement (watch stimuli)
Model Summary
Model

R Square Adjusted R Square Std. Error of the Estimate

0.227 a 0.051

0.028

1.07086

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

2.485

2.485

2.167

.149a

Residual

45.870

40

1.147

Total

48.355

41

Model
1

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.36b Experiment 1: Hypothesis 13b: effect of environmental


concern (enduring involvement with the environment) on
attitude towards the ad (watch stimuli)
Model Summary
Model
1

R Square Adjusted R Square Std. Error of the Estimate

.0.157a 0.025

0.000

1.02015

a. Predictors: (Constant), TOTAL_ENV_CONCERN

166

Table 6.36b (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

1.052

1.052

1.009

.321a

Residual

41.702

40

1.043

Total

42.754

41

Model
1

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: ATTITUDE_AD

Table 6.36c Experiment 1: Hypothesis 13c: effect of environmental


concern (enduring involvement with the environment) on
purchase intention (watch stimuli)
Model Summary
Model
1

R
0.233

R Square Adjusted R Square Std. Error of the Estimate


a

0.054

0.031

1.51735

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

5.284

5.284

2.295

.138a

Residual

92.094

40

2.302

Total

97.378

41

Model
1

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: PURCHASE_INTENTION

H14 was supported as message involvement significantly predicted


attitude towards ad ( =0.716, t(39)=6.482, p<0.001). Model fit was good as
R2=0.51. Table 6.37 shows the results.

167

Table 6.37

Experiment 1: Hypothesis 14: effect of message involvement


on attitude towards ad (watch stimuli)
Model Summary

Model
1

R
0.716

R Square Adjusted R Square Std. Error of the Estimate


a

0.512

0.500

0.72201

a. Predictors: (Constant), MESSAGE_INVOLVEMENT


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

21.902

21.902

42.015

.000a

Residual

20.852

40

.521

Total

42.754

41

Model
1

a. Predictors: (Constant), MESSAGE_INVOLVEMENT


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Model
1 (Constant)
MESSAGE_INVOLVEMENT

Unstandardized
Coefficients
B

Std. Error

2.096

.515

.673

.104

Standardized
Coefficients

Sig.

Beta
4.068 .000
.716

6.482 .000

a. Dependent Variable: ATTITUDE_AD

H15 and H16 were also supported as attitude towards the ad


significantly predicted the attitude towards the brand ( =0.709, t(39)=6.367,
p<0.001, R2=0.50) and attitude towards the brand significantly predicted the
purchase intention ( =0.744, t(39)=7.037, p<0.001, R2=0.55). Tables 6.38 and
6.39 show the regression results for H15 and H16.

168

Table 6.38

Experiment 1: Hypothesis 15: effect of attitude towards ad


on attitude towards brand (watch stimuli)
Model Summary

Model
1

R
0.709

R Square Adjusted R Square Std. Error of the Estimate


a

0.503

0.491

0.90473

a. Predictors: (Constant), ATTITUDE_AD


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

33.185

33.185

40.542

.000a

Residual

32.741

40

.819

Total

65.926

41

Model
1

a. Predictors: (Constant), ATTITUDE_AD


b. Dependent Variable: ATTITUDE_BRAND
Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

.503

.754

ATTITUDE_AD

.881

.138

Standardized
Coefficients

Sig.

.666

.509

6.367

.000

Beta
.709

a. Dependent Variable: ATTITUDE_BRAND

Table 6.39

Experiment 1: Hypothesis 16: effect of attitude towards


brand on purchase intention (watch stimuli)
Model Summary

Model R
1

0.744

R Square Adjusted R Square Std. Error of the Estimate


a

0.553

0.542

a. Predictors: (Constant), ATTITUDE_BRAND

1.04296

169

Table 6.39 (Continued)


ANOVAb
Sum of
Squares

Model
1

Mean
Square

df

Regression

53.868

53.868

Residual

43.511

40

1.088

Total

97.378

41

F
49.521

Sig.
.000a

a. Predictors: (Constant), ATTITUDE_BRAND


b. Dependent Variable: PURCHASE_INTENTION
Coefficientsa
Model

Unstandardized
Coefficients
B

1 (Constant)
ATTITUDE_BRAND

Standardized
Coefficients

Std. Error

.097

.690

.904

.128

Sig.

Beta
.141
.744

.889

7.037 .000

a. Dependent Variable: PURCHASE_INTENTION

Table 6.40 shows the summary of the results regarding the


relationship between the PMT variable, involvement, attitudes and purchase
intention for the watch stimuli.
Table 6.40

Hypothesis
H9a

Experiment 1: Summary of hypotheses tests regarding


relationship between PMT variables, involvement, attitudes
and intentions (H9 H16) (watch stimuli)
Predictor

Perceived
Severity and

Dependent
variable

R2

Adjusted Unstandardised Standardised


R2
coefficient B coefficient

Fear

Not
Significant

Response
Efficacy

Not
Significant

Perceived
vulnerability
H9a

Perceived
Severity and
Perceived
vulnerability

170

Table 6.40 (Continued)


Hypothesis
H9c

H10a

Predictor
Perceived
Severity and
Perceived
vulnerability
Perceived
Severity
Perceived
Vulnerability
Fear

Dependent
variable
Self Efficacy

R2

Adjusted Unstandardised Standardised


R2
coefficient B coefficient
Not Tested

Message
0.135 0.113
Involvement
H10b
Message
0.29 0.27
Involvement
H10c
Message
Involvement
H10d
Response
Message
0.15 0.14
Efficacy
Involvement
H10e
Self Efficacy Message
Involvement
H11a
Response
Attitude
efficacy
towards ad
H11b
Self Efficacy Attitude
towards ad
H11c
Response
Purchase
efficacy
Intention
H11d
Self Efficacy Purchase
Intention
H12a
Environmental Perceived
Knowledge
Severity
H12a
Environmental Perceived
Knowledge
Vulnerability
H12a
Environmental Fear
Knowledge
H12b
Environmental Message
Knowledge
Involvement
H13a
Environmental Message
concern
Involvement
H13b
Environmental Attitude
concern
towards ad
H13c
Environmental Purchase
concern
Intention
H14
Message
Attitude
0.51 0.50
Involvement towards ad
H15
Attitude
Attitude
0.50 0.49
towards ad
towards
brand
H16
Attitude
Purchase
0.55 0.54
towards brand Intention
***p <.001 **p <.01 *p <.05; n= 41

0.433

0.367*

0.41

0.54***

0.345

Not
Significant
0.39*
Not tested
Not
Significant
Not tested
Not
Significant
Not tested

0.673

Not
Significant
Not
Significant
Not
Significant
Not
Significant
Not
Significant
Not
significant
Not
significant
0.716***

0.88

0.70***

0.90

0.74***

171

6.3.7

Conclusions from Experiment 1


This experiment examined the effects of temporal and geographical

framing of threat on PMT variables and the subsequent effects of the PMT
variables on message involvement, attitudes and purchase intention.
Although the hypothesized relationship regarding the main effects were not
significant, it was seen that temporal proximity of the threat had a significant
effect on perceived severity in the case of mobile phone stimuli. This finding
supports current research that argue that threats in day terms are considered
more closer that those that are presented in year terms (Chandran & Menon
2004). It also shows that the self-positivity bias is reduced (Gilovich et al
1993; Raghubir & Menon 1998) as people find a temporally closer threat
relevant. However this effect was not observed with the watch stimuli. This
could be because plastic waste pollution was viewed as a severe threat by
most participants as the mean value of perceived severity was very high
(6.39) when compared to the problem of e-waste (5.9). Plastic waste can be a
more familiar issue in India as the consumer encounters regular mandatory
governmental instructions and news articles on this issue. Hence for a familiar
issue, perceived severity is rated high when compared to an unfamiliar issue
like e-waste. This difference in the arousal of fear based on issue familiarity
has been discussed by Pelsmacker et al (2011). However, only perceived
severity was viewed differently in this experiment. In a similar vein,
Obermiller (1995) also found that different appeals worked for familiar and
unfamiliar issues. The reported perceived vulnerability and fear were almost
similar in both the cases. This could be again because of issue familiarity as
the watch stimulus highlights the threat of plastic waste. Therefore consumers
are more aware of the issue of plastic waste when compared to e-waste. Of
late, the government of India mandates the pricing of plastic bags that are
used to packing the goods sold by a number of retail stores. The stores also
prominently display statutory messages advocating the reduction of plastic.

172

Therefore, for a familiar issue, perceived severity and perceived vulnerability


could have had a greater impact when compared to an unfamiliar issue like ewaste disposal.
The hypothesized effect of CFC was not supported. However, in
the case of mobile phone stimulus, CFC interacted with the factors to produce
significant effects on perceived severity and perceived vulnerability. This
finding supports existing literature that consumers perception of risk varied
based on their temporal orientation (Orbell et al 2004 & Orbell & Hagger
2006; Morison et al 2010). There were no statistically significant interaction
effects of temporal proximity and geographical proximity of the threat on the
PMT variables for both the watch and mobile phone stimuli. Most participants
exhibited high levels of perceived severity, vulnerability and fear towards
environmental threats.
In both the cases, perceived vulnerability and response efficacy
significantly predicted message involvement. This is similar to the finding by
Cauberghe et al (2009). Response efficacy is a variable that is linked to
Perceived consumer effectiveness (PCE) and is shown to be related to
consumers environmental behaviour (Gilg et al 2005).
Unlike previous studies environmental knowledge did not predict
severity or fear as hypothesized. In the case of watch stimuli, environmental
knowledge predicted perceived severity in the positive direction contrary to
the hypothesized nature of the relationship, but in the case of mobile phone
stimuli environmental knowledge did not have any effect.
Most significantly, environmental concern did not have any effect
on attitudes, message involvement or purchase intention. This is in direct
contrast to the propositions put forward by the ELM (Petty & Cacioppo 1986).
Issue involvement is supposed to activate message elaboration and therefore

173

increase involvement with the message (Maheswaran & Meyers-Levy 1990).


However consumers who reported high environmental concern (issue
involvement) did not exhibit this behaviour. This highlights the fact that better
measures are needed to assess environmental concern or enduring
involvement with the environment. Most respondents probably report
environmental concern as a socially desirable response (Bord et al 1998;
Ewert & Baker 2001; Ewert & Galloway 2009) and hence the results did not
support the related hypothesis.
Attitude towards the advertisement significantly predicted attitude
towards the brand and attitude towards the brand significantly predicted
purchase intentions in both the scenarios. This is in congruence with the dual
mediation hypothesis that states that attitude related cognitions affect the
attitude towards the ad, brand and in turn behaviour related to purchase
intentions (MacKenzie et al 1986; Teng et al 2007). The results support the
findings from most advertising studies that show that attitude towards
advertising has a strong influence on attitude towards the brand under high
involvement conditions (Gardner 1985; Park & Young 1986; Muehling &
Laczniak 1988).
This experiment showed that response efficacy and perceived
vulnerability

greatly

increased

message

involvement

and

message

involvement subsequently influenced attitudes and intentions. However


perceived severity and vulnerability did not cause fear arousal in both the
scenarios. Therefore it is necessary to investigate if fear arousal can be
obtained using a different message frame. Hence, study 2 was conducted
using goal frames and threat level as factors to increase the success of the
manipulations. Although it is not essential to conduct the two studies (study1
and study2) sequentially, they were conducted one after another in this
research to examine the effect of the framing manipulations.

174

6.4

STUDY 2: EXPERIMENT 2: THREAT LEVELS AND GOAL


FRAMING (WATCH STIMULI)
Experiment 2 was designed to use threat levels and goal frames to

influence the PMT variables. Plastic waste seemed to generate higher levels
of scores for the threat appraisal variables in the previous study. Therefore the
experiment was conducted with wristwatch as the chosen product. The
advertisements were designed for this product (Ad3). The results of this
experiment were used to design the stimuli for the next experiment.
6.4.1

Experimental Design
A 2 (threat level: high vs. low) x 2 (goal frame: loss vs. gain)

between subjects experimental design was utilized to investigate the


hypotheses. This resulted in four possible combinations of the stimuli. Sixty
nine postgraduate M.E. students from a large South Indian University (95.7 %
male, median age=22) were randomly assigned to the four possible conditions
for the watch stimuli.
The experimental procedure was the same as Experiment 1. Data
collection was through a paper and pencil questionnaire. Students first filled
the questionnaire (Q3) containing dependent variables (Appendix 5). Next,
they were asked to answer the filler task which asked them to list the reason
why they liked their favourite celebrity similar to Experiment 1. On
completion, they filled counterbalanced questionnaires on environmental
concern and objective environmental knowledge. This was similar to the
questions in Experiment 1. The personality variable CFC was not included in
this questionnaire as it was related only to temporal framing.

175

6.4.2

Stimuli
A total of four print advertisements were developed for the four

cells: high threat level and loss frame, high threat level and gain frame, low
threat level and loss frame and low threat level and gain frame. The
advertisements also listed the environment friendly features of the watch.
In the low threat conditions, the advertisement highlighted the fact
that burning plastic waste may cause various health problems. In the high
threat condition, the threat was specific and vivid language was used to
indicate that toxins from burning waste may cause cancer or respiratory
problems. The loss frame mentioned that choosing plastic products will
accelerate air pollution and increase the chances of health hazards caused by
pollution. The gain frame emphasized that by choosing a green product one
can slow down air pollution and reduce the chances of health hazards caused
by pollution. The watch advertisement contained further details about its
biodegradability.
6.4.3

Treatment Validity
The four print advertisements were analyzed by an expert panel to

assess if it contained the necessary variations in the threat level and message
frames. This panel consisted of 3 marketing professors. The changes
suggested by the panel were made and the final versions of the advertisements
are shown in Appendix 6 (Figure A6.1, Figure A6.2, Figure A6.3, and Figure
A6.4).
6.4.4

Manipulation Checks
Two questions were included in the questionnaire to check the

manipulations. Threat level manipulations were checked by including a

176

multiple response question: What health problems did the advertisement


highlight? Choose only one answer. The options were (1) Diseases (2)
Cancer (3) Respiratory problems (4) Cancer and respiratory problems. Frame
manipulations were checked asking the respondents to rate the following
questions: I can gain health benefits by buying biodegradable products, I
can lose important health benefits if I dont buy biodegradable products. The
response to these items was measured using seven point Likert scales
anchored from 1 = Strongly Disagree and 7 = Strongly Agree.
6.4.5

Modified Dependent Variables


The study used the same dependent variables as Experiment 1.

However some of the PMT variables were modified to improve their


reliabilities. Perceived severity and self-efficacy variables were modified as
the reliabilities were low in Experiment1. Response efficacy and perceived
vulnerability were also changed to reflect the changes in the independent
factors. The variables were now adapted from the risk behaviour diagnosis
scale (Witte et al 1996), as this scale is widely used in measuring risks
associated with health messages. Apart from these changes, since the threat
levels varied in the advertisements, efficacy variables were changed to
measure generic diseases rather than specifying particular diseases like
respiratory diseases and cancer. The changed variables are described below
and the entire questionnaire (Q3) is shown in Appendix 5.
Perceived severity
Perceived severity was measured using a three item seven point
scale where 1 = Strongly Disagree and 7 = Strongly Agree. Participants were
asked to indicate their responses on the following statements: I believe that
plastic waste pollution is a serious threat to human health, I believe that

177

plastic waste disposal may cause severe health issues., I believe that plastic
waste pollution is extremely harmful.
Perceived vulnerability
Perceived vulnerability was measured using a three item seven
point scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine
the participants perceived susceptibility to the threat. Participants were asked
to indicate their responses on the following statements. It is possible that I
might get affected by diseases caused by plastic waste pollution., It is
probable that I will suffer from various diseases caused by plastic waste
pollution., I am at risk for getting health problems caused by plastic waste
pollution. These items were collapsed into a single perceived vulnerability
score.
Response efficacy
Response efficacy was measured using a three item seven point
scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine
whether participants believed if purchasing biodegradable products averted
the threat. Participants rated their responses on the following statements:
Purchasing biodegradable products is a highly effective way of preventing
diseases due to plastic pollution, Buying biodegradable products will
significantly lower my risk of being affected by diseases caused by plastic
pollution, Buying biodegradable products is an effective method of
reducing threats caused by plastic pollution to human health. These items
were combined into a single response efficacy score.

178

Self efficacy
Self efficacy was measured using a three item seven point scales
where 1 = Strongly Disagree and 7 = Strongly Agree to determine whether
participants believed if they were capable of averting the threat. Participants
rated their responses on the following statements: I am capable of identifying
and purchasing biodegradable products, I can easily switch over to
biodegradable products to prevent future health problems, It is not difficult
for me to check if products contain plastic or not.
6.4.6

Results of Experiment 2 - Threat Levels and Goal Framing


This study was conducted with the watch stimuli to evaluate the

effect of different threat levels (low/high) and goal frames (gain vs. loss) on
the PMT variables. The effect of PMT variables on involvement and the
subsequent influence of involvement on attitudes and purchase intention were
also evaluated.
6.4.6.1

Manipulation checks
If the participant under high threat condition chose any other

answer, apart from the generic diseases, the manipulation was considered
successful. A chi-square test, comparing the observed frequencies of cases
with the correct evaluation of the threat with the expected frequencies,
revealed that the threat manipulation was successful only in the high threat
condition. The results can be seen in Table 6.41 below. The threat condition
that was assigned to them was correctly identified by 94.2% of the
participants in the high threat condition. This showed that the manipulation
worked for the high threat condition. Even under low threat condition

179

participants viewed the threat as high (64.7% of them viewed the threat as
high).
Table 6.41 Experiment 2: Manipulation check for threat levels
Threat_level * MC_THREAT Crosstabulation
Count

Threat_level

high
low

Total

Value
Pearson ChiSquare

5.515a

MC_THREAT
Cancer and Resp
Generic
31
4
22
53

12
16

Total
35
34
69

Chi-Square Tests
df Asymp. Sig. (2- Exact Sig. (2- Exact Sig. (1sided)
sided)
sided)
1
.019

Continuity
4.256
1
.039
Correctionb
Likelihood Ratio
5.707
1
.017
Fisher's Exact
.024
.019
Test
N of Valid Cases
69
a. 0 cells (.0%) have expected count less than 5. The minimum expected count is
7.88. b. Computed only for a 2x2 table

The frame manipulation was not successful as there was no


significant main effect of the frame manipulation (loss vs. gain) in both the
conditions (Table 6.42).
Similar to Experiment 1, unsuccessful manipulation checks were
not of great concern and did not indicate that the manipulation of the
independent variable failed (Sigall & Mills 1998). Hence further analyses
were conducted.

180

Table 6.42 Experiment 2: Manipulation check for frame type


ANOVA
Sum of Squares df Mean Square F
MC_GAIN Between Groups 1.503
1 1.503
.535
Within Groups 188.265
Total
189.768
MC_LOSS Between Groups 4.304

67 2.810
68
1 4.304

Within Groups
Total

67 1.915
68

6.4.6.2

128.333
132.638

Sig.
.467

2.247 .139

Scale reliability
The internal consistency of the scales was assessed using Cronbach

. The Table 6.43 below shows the reliability scores.


Table 6.43 Experiment 2: Reliability scores using watch stimulus
Construct

Cronbach

Perceived severity

0.84

Perceived vulnerability

0.71

Response Efficacy

0.64

Self Efficacy

0.56

Message involvement

0.80

Fear

0.92

Attitude towards ad

0.86

Attitude towards brand

0.91

Purchase intention

0.91

Environmental concern

0.78

All the variables except self-efficacy had reliability scores exceeding 0.6.
Self-efficacy also has adequate reliability in this case. The results suggest that

181

the instrument was reasonably reliable. The scale reliability for perceived
severity increased with the revision.
6.4.6.3

Hypotheses tests of the effect of manipulations on PMT


variables
Tables 6.44 and 6.45 show the distribution characteristics and the

group wise means of the protection motivation variables. It can be seen that
the average values for perceived severity and perceived vulnerability are high
and closer to the maximum score. The group-wise means also show that there
are not much variations in the scores across the groups.
Table 6.44

Experiment 2: Distribution characteristics of the protection


motivation variables (watch stimuli)
Minimum Maximum Mean Std. Deviation

PERC_SEV

1.00

7.00

6.22

1.00

PERC_VUL

2.33

7.00

5.31

1.20

RESP_EFFICACY

3.33

7.00

5.92

0.82

SELF_EFFICACY

2.00

6.67

4.88

1.26

FEAR

1.00

7.00

4.34

1.60

Table 6.45

Experiment 2: Group wise mean values of protection


motivation variables for the (watch stimuli)
Self

Perceived
Severity

Perceived
Vulnerability

Response
Efficacy

Efficacy

Threat level: High

6.25

5.32

5.90

4.91

4.25

Threat level: Low

6.17

5.29

5.94

4.85

4.42

Goal frame: Gain

6.27

5.43

6.07

5.06

4.49

Goal frame: Loss

6.16

5.19

5.78

4.72

4.20

Factor

Fear

182

To test the hypotheses regarding the effect of manipulations on the


PMT variables, MANOVA was conducted to test the hypotheses. Hypothesis
5 stated that participants who viewed advertisements with higher threats
levels would report higher severity and vulnerability when compared to
consumers who viewed weaker threats. A one-way MANOVA was conducted
to test this hypothesis. The one-way MANOVA results were: Pillais
Trace=0.002; Wilks lambda = 0.998; Hotellings Trace and Roys Largest
Root = 0.002, F(2,66)=0.057, p > 0.05) (Table 6.46a). The results indicate
that there was no statistically significant difference in severity and
vulnerability based on threat levels and hence Table 6.46b was not further
interpreted. Therefore hypothesis 5 (H5) was not supported.
Table 6.46a Experiment 2: Hypothesis 5: multivariate tests (watch
stimuli)
Multivariate Testsb
Effect
Intercept

Value

Hypothesis df Error df Sig.

.977 1387.169a

2.000

66.000 .000

.023 1387.169

2.000

66.000 .000

42.035 1387.16 a

2.000

66.000 .000

Roy's Largest Root 42.035 1387.16 9a

2.000

66.000 .000

Pillai's Trace
Wilks' Lambda
Hotelling's Trace

Threat_level Pillai's Trace


Wilks' Lambda
Hotelling's Trace

.002

.057a

2.000

66.000 .945

.998

2.000

66.000 .945

2.000

66.000 .945

2.000

66.000 .945

.002

Roy's Largest Root .002


a. Exact statistic
b. Design: Intercept + Threat_level

.057
.057

.057

183

Table 6.46b Experiment 2: Hypothesis 5: tests of between-subjects effects


(watch stimuli)
Tests of Between-Subjects Effects
Dependent
Variable

Type III Sum


of Squares

df

Mean
Square

Sig.

Corrected
Model

PERC_SEV

.112a

.112

.111

.740

PERC_VUL

.015

.015

.010

.919

Intercept

PERC_SEV

2666.199

2666.199

2641.484 .000

PERC_VUL

1944.363

1944.363

1340.709 .000

PERC_SEV

.112

.112

.111

.740

PERC_VUL

.015

.015

.010

.919

PERC_SEV

67.627

67

1.009

PERC_VUL

97.167

67

1.450

PERC_SEV

2735.000

69

PERC_VUL

2042.111

69

PERC_SEV

67.739

68

Source

Threat_level
Error
Total
Corrected
Total

a. R Squared = .002 (Adjusted R Squared = -.013)


b. R Squared = .000 (Adjusted R Squared = -.015)

The results also indicate that there was no statistically significant


difference in severity and vulnerability based on frame type (Pillais
Trace=0.010; Wilks lambda = 0.990; Hotellings Trace and Roys Largest
Root = 0.010, F(2,66)=0.341, p > 0.05). Therefore hypothesis 6 (H6) was not
supported. The following tables (Tables 6.47a and 6.47b) show the results.
Table 6.47a Experiment 2: Hypothesis 6:
stimuli)

multivariate tests (watch

Multivariate Testsb
Effect
Intercept Pillai's Trace
Wilks' Lambda

Hypothesis Error
df
df

Value

.977

1392.001a

2.000

66.000

.000

.023

2.000

66.000

.000

1392.001

Sig.

184

Table 6.47a (Continued)


Effect

Hypothesis
df
a
42.182 1392.001
2.000
a
42.182 1392.001
2.000
a
.010
.341
2.000
a
.990
.341
2.000
a
.010
.341
2.000
a
.010
.341
2.000
Value

Hotelling's Trace
Roy's Largest Root
frame
Pillai's Trace
Wilks' Lambda
Hotelling's Trace
Roy's Largest Root
a. Exact statistic
b. Design: Intercept + frame

Error
df
66.000
66.000
66.000
66.000
66.000
66.000

Sig.
.000
.000
.712
.712
.712
.712

Table 6.47b Experiment 2: Hypothesis 6: tests of between-subjects effects


(watch stimuli)
Tests of Between-Subjects Effects
Source

Dependent
Variable

Type III Sum of


df
Squares

Mean
Square

Sig.

Corrected
Model

PERC_SEV

.194a

.194

.192

.663

PERC_VUL

.991

.991

.690

.409

Intercept

PERC_SEV

2664.194

2664.194 2642.679 .000

PERC_VUL

1945.068

1945.068 1354.799 .000

PERC_SEV

.194

.194

.192

.663

PERC_VUL

.991

.991

.690

.409

PERC_SEV

67.545

67

1.008

PERC_VUL

96.191

67

1.436

PERC_SEV

2735.000

69

PERC_VUL

2042.111

69

PERC_SEV

67.739

68

PERC_VUL

97.182

68

frame
Error
Total
Corrected
Total

a. R Squared = .003 (Adjusted R Squared = -.012)


b. R Squared = .010 (Adjusted R Squared = -.005)

It can be seen from Tables 6.48a and 6.48b that the proposed
interaction between threat levels and frames was also not supported (Pillais

185

Trace=0.079; Wilks lambda = 0.921; Hotellings Trace and Roys Largest


Root = 0.086 (F2,64)=2.539, p>0.05). Therefore hypothesis 7 (H7) was not
supported.
Table 6.48a Experiment 2: Hypothesis 7:
stimuli)

multivariate tests (watch

Multivariate Testsb
Effect
Intercept

Pillai's Trace
Wilks' Lambda

frame

64.000 .000

Hotelling's Trace 43.124 1379.959a 2.000

64.000 .000

Roy's Largest
Root

43.124 1379.959a 2.000

64.000 .000

Pillai's Trace

.010

.325a

2.000

64.000 .724

.990

.325

2.000

64.000 .724

Hotelling's Trace .010

.325

2.000

64.000 .724

Roy's Largest
Root

.010

.325a

2.000

64.000 .724

Pillai's Trace

.002

.061a

2.000

64.000 .941

.998

.061

2.000

64.000 .941

Hotelling's Trace .002

.061

2.000

64.000 .941

Roy's Largest
Root

.002

.061a

2.000

64.000 .941

Pillai's Trace

.074

2.539a

2.000

64.000 .087

.926

2.000

64.000 .087

Hotelling's Trace .079

2.539

2.000

64.000 .087

Roy's Largest
Root

2.539a

2.000

64.000 .087

Wilks' Lambda

.023

1379.959a 2.000

Error
Sig.
df
64.000 .000

Wilks' Lambda

frame *
Threat_level

.977

Hypothesis
df

1379.959 2.000

Wilks' Lambda

Threat_level

Value

.079

2.539

a. Exact statistic
b. Design: Intercept + frame + Threat_level + frame * Threat_level

186

From Table 6.48b it could be seen that the interaction of the independent
factors had an effect on perceived vulnerability. Hence a followup ANOVA
revealed this effect. The ANOVA results are shown in Table 6.48c. Figure
6.2 shows that gain frames and high levels of threat produced higher
vulnerability scores. Under loss frame conditions, low threat produced higher
scores of perceived vulnerability.
Table 6.48b Experiment 2: Hypothesis 7:
stimuli)

multivariate tests (watch

Tests of Between-Subjects Effects


Dependent
Variable

Type III Sum


of Squares

df

Mean
Square

Sig.

Corrected
Model

PERC_SEV

.743a

.248

.240

.868

PERC_VUL

7.784

2.595

1.886

.141

Intercept

PERC_SEV

2661.319

2661.319

2582.032 .000

PERC_VUL

1940.361

1940.361

1410.803 .000

PERC_SEV

.180

.180

.175

.677

PERC_VUL

.907

.907

.659

.420

PERC_SEV

.128

.128

.124

.726

PERC_VUL

.050

.050

.036

.850

frame *
Threat_level

PERC_SEV

.442

.442

.428

.515

PERC_VUL

6.781

6.781

4.930

.030

Error

PERC_SEV

66.996

65

1.031

PERC_VUL

89.398

65

1.375

PERC_SEV

2735.000

69

PERC_VUL

2042.111

69

PERC_SEV

67.739

68

PERC_VUL

97.182

68

Source

Frame

Threat_level

Total
Corrected
Total

a. R Squared = .011 (Adjusted R Squared = -.035)


b. R Squared = .080 (Adjusted R Squared = .038)

187

Table 6.48c Experiment 2: Hypothesis 7: interaction effect of goal


frames and threat levels on perceived vulnerability
Tests of Between-Subjects Effects
Dependent Variable:PERC_VUL
Source
Corrected
Model
Intercept

Type III
Mean
Sum of df
Square
Squares

Partial
Noncent. Observed
Eta
Parameter Powerb
Squared

Sig.

1.886

.141

.080

5.659

.467

1940.361 1 1940.361 1410.803 .000

.956

1410.803

1.000

7.784a

2.595

Threat_level

.050

.050

.036

.850

.001

.036

.054

frame
Threat_level
* frame

.907

.907

.659

.420

.010

.659

.126

6.781

6.781

4.930

.030

.071

4.930

.590

89.398

65

1.375

Error
Total

2042.111 69

Corrected
97.182 68
Total
a. R Squared = .080 (Adjusted R Squared = .038) b. Computed using alpha = .05

Figure 6.2

Experiment 2 effect of threat levels and goal on perceived


vulnerability

188

In order to examine H8a regression analyses was done with three


predictors: framing, environmental concern and an interaction term of these
variables with purchase intention as the dependent variable. Framing was
dummy coded with the loss-frame message condition allocated a value of 0
and the gain-frame message condition a value of 1.
The interaction terms were calculated as a product of frame type
and environmental concern (frame x environmental concern). The hypothesis
was not supported as the interaction between the variables did not predict
purchase intention. Since the model was significant, a follow up stepwise
regression revealed that only environmental concern predicted purchase
intention. Table 6.49a shows this interaction.
Table 6.49a Experiment 2: Hypothesis 8a: interaction of frame and
environmental concern on purchase intention (watch
stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.444a

0.197

0.147

1.25387

a. Predictors: (Constant), TOTAL_ENV_CONCERN, FRAME_CODED,


ENV_CONC_X_FRAME
ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

21.394

7.131

4.461

.007a

Residual

103.901

65

1.598

Total

125.295

68

Model
1

a. Predictors: (Constant), TOTAL_ENV_CONCERN, FRAME_CODED,


ENV_CONC_X_FRAME
b. Dependent Variable: PURCHASE_INTENTION

189

Table 6.49a (Continued)


Coefficientsa
Unstandardized
Coefficients

Standardized
Coefficients

Model
B
1 (Constant)

Std.
Error

Sig.

-.146

.884

Beta

-.688

4.717

FRAME_CODED

.127

3.294

.047

.039

.969

ENV_CONC_X_FRAME

.007

.523

.018

.014

.989

TOTAL_ENV_CONCERN .886

.750

.401

1.181 .242

a. Dependent Variable: PURCHASE_INTENTION

Similarly H8b was also not supported as message involvement did


not interact with frame type to produce an effect on purchase intentions.
Table 6.49b shows this interaction. Since the model was significant, a follow
up stepwise regression revealed that only message involvement significantly
predicted purchase intention.
Table 6.49b Experiment 2: Hypothesis 8b: interaction of frame and
message involvement on purchase intention (watch stimuli)
Model Summary
Model
1

R
.0.515a

R Square

Adjusted R Square

0.266

0.232

Std. Error of the


Estimate
1.18986

Predictors: (Constant), MESS_INV_X_FRAME, MESSAGE_INVOLVEMENT,


FRAME_CODED

190

Table 6.49b (Continued)


ANOVAb
Model
1

Sum of
Squares

Mean
Square

df

Regression 33.270

11.090

Residual

92.025

65

1.416

Total

125.295

68

F
7.833

Sig.
.000a

a. Predictors: (Constant), MESS_INV_X_FRAME,


MESSAGE_INVOLVEMENT, FRAME_CODED
b. Dependent Variable: PURCHASE_INTENTION
Coefficientsa
Model

Unstandardized
Coefficients

Standardized
Coefficients

Sig.

Std. Error

3.679

2.110

-1.331

1.470

-.493

-.905 .369

MESSAGE_INVOLVEMENT

.254

.400

.204

.634 .528

MESS_INV_X_FRAME

.270

.276

.637

.979 .331

1 (Constant)
FRAME_CODED

Beta
1.744 .086

a. Dependent Variable: PURCHASE_INTENTION

H8c was not tested as the sample consisted of mostly male subjects
(95.7%). Therefore gender based variations could not be investigated.
The summary of results is shown in Table 6.50. The table clearly
highlights the fact that the factors were not successful in producing the
hypothesized main effects or interaction effects.

191

Table 6.50

Experiment 2: Summary of the effect of manipulations on


PMT variables with the watch stimulus

Hypothesis

Factor

Perceived
severity

Perceived
vulnerability

H5

Threat level

H6

Goal frame

H7

Threat level * Goal Frame

Purchase
Intention
H8a

Environmental concern *
Goal Frame

H8b

Message Involvement * Goal


Frame

6.4.6.4

Hypotheses tests of the relationship among PMT variables,


involvement, attitudes and intentions
Similar to Experiment 1, standard simple or multiple linear

regression (ordinary least squares (OLS)) analyses were run to ascertain the
effect of the predictor variables. Only one or two predictors were considered
at

time.

Similarly

assumptions

regarding

like

multicollinearity,

homoscedascity and linearity were met in all the scenarios.


A multiple regression analysis was conducted to evaluate how well
perceived severity and perceived vulnerability predicted fear. H9a was
supported as both perceived severity ( =0.318, t(66)=2.622, p<0.05) and
perceived vulnerability ( =0.273, t(66)=2.254, p<0.05) significantly predicted
fear. The model fit was also good (R2=0.261).

192

Perceived severity also predicted both the efficacy variables.


Tables 6.51a, 6.51b and 6.51c show the regression results.
Table 6.51a Experiment 2: Hypothesis 9a: effect of threat appraisal
components on fear (watch stimuli)
Model Summary
Model

.510a

R Square Adjusted R Square Std. Error of the Estimate


0.261

0.238

1.39272

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

45.110

22.555

11.628

.000a

Residual

128.018

66

1.940

Total

173.128

68

Model
1

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: FEAR
Coefficientsa
Unstandardized
Coefficients

Model

B
1

Std. Error

(Constant)

-.755

1.099

PERC_SEV

.508

.194

PERC_VUL .365

.162

a. Dependent Variable: FEAR

Standardized
Coefficients

Sig.

Beta
-.687

.495

.318

2.622

.011

.273

2.254

.028

193

Table 6.51b Experiment 2: Hypothesis 9b: effect of threat appraisal


components on response efficacy (watch stimuli)
Model Summary
Model R
1

.382

R Square Adjusted R Square Std. Error of the Estimate


a

0.146

0.120

0.77197

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

6.700

3.350

5.621

.006a

Residual

39.332

66

.596

Total

46.032

68

Model
1

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: RESP_EFFICACY
Coefficientsa
Model
1

Unstandardized
Coefficients

Standardized
Coefficients

Sig.

6.497

.000

Std. Error

Beta

(Constant)

3.959

.609

PERC_SEV

.196

.107

.237

1.821

.073

PERC_VUL

.141

.090

.205

1.569

.121

a. Dependent Variable: RESP_EFFICACY

From Table 6.51b it can be seen that although the model was
significant, the predictors were not significant. Hence H9b was not supported.
A stepwise regression was conducted to evaluate which one of the variables
contributed to the model. The results showed that perceived severity alone
predicted response efficacy as shown in Table 6.51 b1 (R2=0.114 =0.337,
t(66)=2.931, p<0.01).

194

Table 6.51b1 Experiment 2: Effect of threat appraisal components on


response efficacy (watch stimuli)
Model Summary
Model
1

R
.337

R Square Adjusted R Square Std. Error of the Estimate


a

0.114

0.100

0.78036

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

5.232

5.232

8.592

.005a

Residual

40.800

67

.609

Total

46.032

68

Model
1

a. Predictors: (Constant), PERC_SEV


b. Dependent Variable: RESP_EFFICACY
Coefficientsa
Model
1

Unstandardized
Coefficients

Standardized
Coefficients

Std. Error

(Constant)

4.195

.597

PERC_SEV

.278

.095

Sig.

7.027

.000

2.931

.005

Beta
.337

a. Dependent Variable: RESP_EFFICACY


Excluded Variablesb
Model
1

PERC_VUL

Beta In
.205

t
1.569

Sig.
.121

a. Predictors in the Model: (Constant), PERC_SEV


b. Dependent Variable: RESP_EFFICACY

Partial
Correlation
.190

Collinearity
Statistics
Tolerance
.762

195

The results showed that perceived severity alone signficantly


predicts self efficacy as shown in Table 6.51c (R2=0.090

=0.331,

t(66)=2.459, p<0.01). Hence H9c was not supported.


Table 6.51c Experiment 2: Hypothesis 9c: effect of threat appraisal
components on self efficacy (watch stimuli)
Model Summary
Model

.300a

R Square Adjusted R Square Std. Error of the Estimate


0.090

0.062

1.22004

a. Predictors: (Constant), PERC_VUL, PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

9.721

4.861

3.265

.044a

Residual

98.240

66

1.488

Total

107.961

68

Model
1

a. Predictors: (Constant), PERC_VUL, PERC_SEV


b. Dependent Variable: SELF_EFFICACY
Coefficientsa
Unstandardized
Coefficients

Model

B
1

(Constant)

2.734

Standardized
Coefficients

Std. Error

Sig.

Beta

.963

2.839

.006

PERC_SEV .418

.170

.331

2.459

.017

PERC_VUL

.142

-.080

-.593

.555

-.084

a. Dependent Variable: SELF_EFFICACY

Simple regression analyses were run with each of the PMT


variables and each one of them except self-efficacy significantly predicted
message

involvement.

Perceived

severity

( =0.261

t(67)=2.213,

196

p<0.05,R2=0.06),

perceived

vulnerability

( =0.334

t(67)=2.898,

p<0.001,R2=0.111), fear ( =0.554 t(67)=5.447, p<0.001,R2=0.307),

and

response efficacy ( =0.458 t(67)=4.221, p<0.05,R = 0.210) significantly


predict message involvement. The results are shown in the following tables
Tables 6.52a, 6.52b, 6.52c and 6.52d. Therefore H10a, H10b, H10c, H10d
were supported and H10e was not supported.
Table 6.52a Experiment 2: Hypothesis 10a: effect of perceived severity
on message involvement (watch stimuli)
Model Summary
Model
1

R
.261

R Square Adjusted R Square Std. Error of the Estimate


a

0.068

0.054

1.06091

a. Predictors: (Constant), PERC_SEV


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

5.512

5.512

4.898

.030a

Residual

75.411

67

1.126

Total

80.924

68

Model
1

a. Predictors: (Constant), PERC_SEV


b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa
Model
1

Unstandardized
Coefficients
B

Std. Error

(Constant)

3.420

.812

PERC_SEV

.285

.129

Standardized
Coefficients

Sig.

4.214

.000

2.213

.030

Beta
.261

a. Dependent Variable: MESSAGE_INVOLVEMENT

197

Table 6.52b Experiment 2: Hypothesis 10b: effect of perceived


vulnerability on message involvement (watch stimuli)
Model Summary
Model R
1

R Square Adjusted R Square Std. Error of the Estimate


a

.0.334

0.111

0.098

1.03598

a. Predictors: (Constant), PERC_VUL


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

9.016

9.016

8.400

.005a

Residual

71.908

67

1.073

Total

80.924

68

Model
1

a. Predictors: (Constant), PERC_VUL


b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

3.576

.572

PERC_VUL

.305

.105

Standardized
Coefficients

Sig.

6.255

.000

2.898

.005

Beta
.334

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.52c Experiment 2: Hypothesis 10c: effect of fear on message


involvement (watch stimuli)
Model Summary
Model R
1

R Square Adjusted R Square Std. Error of the Estimate

.554a 0.307

0.297

a. Predictors: (Constant), FEAR

0.91495

198

Table 6.52c (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

24.836

24.836

29.668

.000a

Residual

56.088

67

.837

Total

80.924

68

Model
1

a. Predictors: (Constant), FEAR


b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

3.549

.321

FEAR

.379

.070

Standardized
Coefficients

Sig.

11.042

.000

5.447

.000

Beta
.554

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.52d Experiment 2: Hypothesis 10d: effect of response efficacy on


message involvement (watch stimuli)
Model Summary
Model R
1

.458

R Square Adjusted R Square Std. Error of the Estimate


a

0.210

0.198

0.97676

a. Predictors: (Constant), RESP_EFFICACY


ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression 17.001

17.001

Residual

63.922

67

.954

Total

80.924

68

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: MESSAGE_INVOLVEMENT

F
17.820

Sig.
.000a

199

Table 6.52d (Continued)


Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

1.594

.861

RESP_EFFICACY

.608

.144

Standardized
Coefficients

Sig.

1.852

.068

4.221

.000

Beta
.458

a. Dependent Variable: MESSAGE_INVOLVEMENT

Table 6.52e Experiment 2: Hypothesis 10e: effect of self efficacy on


message involvement (watch stimuli)
Model Summary
Model

.236a

R Square Adjusted R Square Std. Error of the Estimate


0.056

0.042

1.06800

a. Predictors: (Constant), SELF_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

4.502

4.502

3.947

.051a

Residual

76.422

67

1.141

Total

80.924

68

Model
1

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa
Model
1

Unstandardized
Coefficients
B

Std. Error

(Constant)

4.196

.518

SELF_EFFICACY

.204

.103

Standardized
Coefficients

a. Dependent Variable: MESSAGE_INVOLVEMENT

Sig.

8.097

.000

1.987

.051

Beta
.236

200

Multiple regression analysis was conducted to evaluate how well


coping variables predicted attitude towards ad and purchase intention. Both
response efficacy ( =0.474, t(67)=4.409, p<0.001, R2=0.225) and self
efficacy ( =0.271, t(67)=2.306, p<0.05,R2=0.074) predicted attitude towards
ad. However they did not predict purchase intention. Therefore H11a and
H11b were supported but H11c and H11d were not. These results can be seen
in Tables 6.53a, 6.53b, 6.53c and 6.53d.
Table 6.53a Experiment 2: Hypothesis 11a: effect of response efficacy on
attitude towards ad (watch stimuli)
Model Summary
Model R
1

R Square Adjusted R Square Std. Error of the Estimate

.474a

0.225

0.213

1.01172

a. Predictors: (Constant), RESP_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

19.897

19.897

19.439

.000a

Residual

68.579

67

1.024

Total

88.477

68

Model
1

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Model
1

Unstandardized
Coefficients
B

Std. Error

(Constant)

1.739

.892

RESP_EFFICACY

.657

.149

a. Dependent Variable: ATTITUDE_AD

Standardized
Coefficients

Sig.

1.950

.055

4.409

.000

Beta
.474

201

Table 6.53b Experiment 2: Hypothesis 11b: effect of self efficacy on


attitude towards ad (watch stimuli)
Model Summary
Model R
1

.271

R Square Adjusted R Square Std. Error of the Estimate


a

0.074

0.060

1.10608

a. Predictors: (Constant), SELF_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

6.508

6.508

5.320

.024a

Residual

81.968

67

1.223

Total

88.477

68

Model
1

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Unstandardized
Coefficients

Model

B
1

(Constant)

4.434

SELF_EFFICACY .246

Standardized
Coefficients

Std. Error

Sig.

Beta

.537
.106

.271

8.261

.000

2.306

.024

a. Dependent Variable: ATTITUDE_AD

Table 6.53c Experiment 2: Hypothesis 11b: effect of response efficacy on


purchase intention (watch stimuli)
Model Summary
Model R
1

.185a

R Square Adjusted R Square Std. Error of the Estimate


0.034

0.020

a. Predictors: (Constant), RESP_EFFICACY

1.34386

202

Table 6.53c (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

4.296

4.296

2.379

.128a

Residual

120.999

67

1.806

Total

125.295

68

Model
1

Regression

a. Predictors: (Constant), RESP_EFFICACY


b. Dependent Variable: PURCHASE_INTENTION

Table 6.53d Experiment 2: Hypothesis 11d: effect of self efficacy on


purchase intention (watch stimuli)
Model Summary
Model R
1

R Square Adjusted R Square Std. Error of the Estimate


a

.0.110

0.012

-0.003

1.35918

a. Predictors: (Constant), SELF_EFFICACY


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

1.521

1.521

.824

.367a

Residual

123.773

67

1.847

Total

125.295

68

Model
1

Regression

a. Predictors: (Constant), SELF_EFFICACY


b. Dependent Variable: PURCHASE_INTENTION

H12a, H12b, H12c and H12d were not supported as environmental


knowledge did not significantly predict perceived severity, perceived
vulnerability, fear or message involvement. Tables 6.54a , 6.54b 6.54c, 6.54d
illustrate the results.

203

Table 6.54a Experiment 2: Hypothesis 12a: effect of environmental


knowledge on perceived severity (watch stimuli)
Model Summary
Model
1

R
0.130

R Square Adjusted R Square Std. Error of the Estimate


a

0.017

0.002

0.99695

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

1.147

1.147

1.154

.286a

Residual

66.592

67

.994

Total

67.739

68

Model
1

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: PERC_SEV

Table 6.54b Experiment 2: Hypothesis 12b: effect of environmental


knowledge on perceived vulnerability (watch stimuli)
Model Summary
Model
1

R
0.092

R Square Adjusted R Square Std. Error of the Estimate


a

0.009

-0.006

1.19921

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Model
1

Sum of
Squares

df

Mean
Square

Regression .829

.829

Residual

96.353

67

1.438

Total

97.182

68

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: PERC_VUL

F
.577

Sig.
.450a

204

Table 6.54c Experiment 2: Hypothesis 12c: effect of environmental


knowledge on fear (watch stimuli)
Model Summary
Model
1

R
0.077

R Square Adjusted R Square Std. Error of the Estimate


a

0.006

-0.009

1.60273

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

1.021

1.021

.398

.530a

Residual

172.107

67

2.569

Total

173.128

68

Model
1

Regression

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: FEAR

Table 6.54d Experiment 2: Hypothesis 12d: effect of environmental


knowledge on message involvement (watch stimuli)
Model Summary
Model
1

R
0.103

R Square Adjusted R Square Std. Error of the Estimate


a

0.011

-0.004

1.09322

a. Predictors: (Constant), ENV_KNOW


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

.850

.850

.711

.402a

Residual

80.073

67

1.195

Total

80.924

68

Model
1

Regression

a. Predictors: (Constant), ENV_KNOW


b. Dependent Variable: MESSAGE_INVOLVEMENT

205

Similar simple regression analysis was conducted to evaluate if


involvement predicted attitude towards ad. Environmental concern was
positively related to message involvement ( =0.457 t(66)=4.203, p<0.001,
R2=0.209) and therefore H13a was supported. H13b was also supported as
environmental concern significantly predicted attitude towards ad ( =0.499
t(66)=4.716, p<0.001, R2=0.249). Environmental concern significantly
predicted purchase intention ( =0.408 t(66)=3.661, p<0.001, R2=0.167) and
therefore H13c was supported. The following tables (Tables 6.55a, 6.55b,
6.55c) show the regression results.
Table 6.55a Experiment 2: Hypothesis 13a: effect of environmental
concern (enduring involvement with the environment) on
message involvement (watch stimuli)

Model
R
1
0.457 a

Model Summary
R Square Adjusted R Square Std. Error of the Estimate
0.209
0.197
0.97766

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Model
1

Regression
Residual

Sum of
Squares
16.883
64.040

Mean
Square
16.883

67

.956

df

Sig.

17.664

.000a

Total
80.924
68
a. Predictors: (Constant), TOTAL_ENV_CONCERN
b. Dependent Variable: MESSAGE_INVOLVEMENT
Coefficientsa

Model

Unstandardized
Coefficients

Standardized
Coefficients

Std.
B
Beta
Error
1 (Constant)
.112
1.215
TOTAL_ENV_CONCERN .811
.193
.457
a. Dependent Variable: MESSAGE_INVOLVEMENT

Sig.

.092 .927
4.203 .000

206

Table 6.55b Experiment 2: Hypothesis 13b: effect of environmental


concern (enduring involvement with the environment) on
attitude towards the ad (watch stimuli)
Model Summary
Model R

R Square Adjusted R Square Std. Error of the Estimate

0.499 a

0.249

0.238

0.99571

a. Predictors: (Constant), TOTAL_ENV_CONCERN


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

22.050

22.050

22.241

.000a

Residual

66.426

67

.991

Total

88.477

68

Model
1

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Unstandardized
Coefficients

Model
1 (Constant)
TOTAL_ENV_CONCERN

Std. Error

-.175

1.237

.927

.197

Standardized
Coefficients

Sig.

Beta
-.141 .888
.499

4.716 .000

a. Dependent Variable: ATTITUDE_AD

Table 6.55c Experiment 2: Hypothesis 13c: effect of environmental


concern (enduring involvement with the environment) on
purchase intention (watch stimuli)
Model Summary
Model

0.408 a

R Square Adjusted R Square Std. Error of the Estimate


0.167

0.154

a. Predictors: (Constant), TOTAL_ENV_CONCERN

1.24834

207

Table 6.55c (Continued)


ANOVAb
Sum of
Squares

Model
1

Mean
Square

df

Regression 20.884

20.884

Residual

104.410

67

1.558

Total

125.295

68

F
13.401

Sig.
.000a

a. Predictors: (Constant), TOTAL_ENV_CONCERN


b. Dependent Variable: PURCHASE_INTENTION
Coefficientsa
Unstandardized
Coefficients

Model

B
1 (Constant)

-.536

TOTAL_ENV_CONCERN .902

Standardized
Coefficients

Std. Error

Sig.

Beta

1.551
.246

-.345 .731
.408

3.661 .000

a. Dependent Variable: PURCHASE_INTENTION

It can be seen from Table 6.56 that H14 was supported as message
involvement predicted attitude towards ad ( =0.718 t(66)=8.452, p<0.001).
The model fit was also good with R2 =0.516.
Table 6.56

Experiment 2: Hypothesis 14: effect of message involvement


on attitude towards ad (watch stimuli)
Model Summary

Model R
1

0.718

R Square Adjusted R Square Std. Error of the Estimate


a

0.516

0.509

a. Predictors: (Constant), MESSAGE_INVOLVEMENT

0.79944

208

Table 6.56 (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

45.657

45.657

71.439

.000a

Residual

42.820

67

.639

Total

88.477

68

Model
1

a. Predictors: (Constant), MESSAGE_INVOLVEMENT


b. Dependent Variable: ATTITUDE_AD
Coefficientsa
Unstandardized
Coefficients

Model

1 (Constant)
MESSAGE_INVOLVEMENT

Std.
Error

1.732

.471

.751

.089

Standardized
Coefficients

Sig.

Beta
3.674 .000
.718

8.452 .000

a. Dependent Variable: ATTITUDE_AD

H15 and H16 were also supported as attitude towards the ad


significantly predicted the attitude towards the brand ( =0.747 t(66)=9.210,
p<0.001, R2=0.559). Table 6.57 illustrates this result. Attitude towards the
brand significantly predicted the purchase intention as seen in Table 6.58
=0.647, t(66)=6.942, p<0.001, R2=0.418).
Table 6.57

Experiment 2: Hypothesis 15: effect of attitude towards ad


on attitude towards brand (watch stimuli)
Model Summary

Model
1

R
0.747

R Square Adjusted R Square Std. Error of the Estimate


a

0.559

0.552

a. Predictors: (Constant), ATTITUDE_AD

0.74046

209

Table 6.57 (Continued)


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

46.512

46.512

84.832

.000a

Residual

36.735

67

.548

Total

83.246

68

Model
1

a. Predictors: (Constant), ATTITUDE_AD


b. Dependent Variable: ATTITUDE_BRAND
Coefficientsa
Unstandardized
Coefficients

Model
1

Std. Error

(Constant)

1.409

.452

ATTITUDE_AD

.725

.079

Standardized
Coefficients

Sig.

3.115

.003

9.210

.000

Beta
.747

a. Dependent Variable: ATTITUDE_BRAND

Table 6.58

Experiment 2: Hypothesis 16: effect of attitude towards


brand on purchase intention (watch stimuli)
Model Summary

Model
1

R
0.647

R Square Adjusted R Square Std. Error of the Estimate


a

0.418

0.410

1.04293

a. Predictors: (Constant), ATTITUDE_BRAND


ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

Regression

52.419

52.419

48.192

.000a

Residual

72.876

67

1.088

Total

125.295

68

Model
1

a. Predictors: (Constant), ATTITUDE_BRAND


b. Dependent Variable: PURCHASE_INTENTION

210

Table 6.58 (Continued)


Coefficientsa
Unstandardized
Coefficients

Model
1

(Constant)
ATTITUDE_BRAND

B
.757
.794

Std. Error
.640
.114

Standardized
Coefficients

Sig.

1.183
6.942

.241
.000

Beta
.647

a. Dependent Variable: PURCHASE_INTENTION

Table 6.59 below shows the summary of the hypotheses tests


regarding the relationship among the PMT variables, involvement, attitudes
and purchase intention.
Table 6.59

Hypothesis
H9a

Experiment 2: Summary of hypotheses tests regarding


relationship between PMT variables, involvement, attitudes
and intentions (H9 H16) (watch stimuli)
Predictor

Perceived
Severity and

Dependent
variable
Fear

R2

Adjusted Unstandardised Standardised


R2
coefficient B coefficient

0.261 0.238

0.508

0.318*

0.365

0.273*

Perceived
vulnerability
H9b

Perceived
Severity and

Response
Efficacy

Not supported
(Only
perceived
severity)

Perceived
vulnerability
H9a

Perceived
Severity and

Self efficacy 0.300 0.090

0.418

0.331*

Perceived
vulnerability
H10a

Perceived
Severity

Message
0.06 0.05
Involvement

0.285

0.261*

H10b

Perceived

Message
0.11 0.09
Involvement

0.305

0.334**

Message
0.30 0.29
Involvement

0.37

0.55***

Vulnerability
H10c

Fear

211

Table 6.59 (Continued)


Hypothesis

Predictor

Dependent
variable

R2

Adjusted Unstandardised Standardised


R2
coefficient B coefficient

H10d

Response
Efficacy

Message
0.21 0.19
Involvement

H10e

Self Efficacy

Message
Involvement

H11a

Response
efficacy

Attitude
towards ad

0.22 0.21

0.65

0.47 ***

H11b

Self Efficacy

Attitude
towards ad

0.07 0.06

0.246

0.271 *

H11c

Response
efficacy

Purchase
Intention

Not
Significant

H11d

Self Efficacy

Purchase
Intention

Not
Significant

H12a

Environmental Perceived
Severity
Knowledge

Not
Significant

H12b

Environmental Perceived
Vulnerabilty
Knowledge

Not
Significant

H12c

Environmental Fear
Knowledge

Not
Significant

H12d

Environmental Message
Involvement
Knowledge

Not
Significant

H13a

Environmental Message
0.20 0.19
concern
Involvement

0.81

0.45***

H13b

Environmental Attitude
concern
towards ad

0.25 0.24

0.92

0.49***

H13c

Environmental Purchase
concern
Intention

0.16 0.15

0.902

0.408***

H14

Message
Involvement

Attitude
towards ad

0.51 0.50

0.751

0.718***

H15

Attitude
towards ad

Attitude
towards
brand

0.55 0.55

0.725

0.747***

H16

Attitude
Purchase
towards brand Intention

0.41 0.41

0.79

0.64***

***p <.001 **p <.01 *p <.05; n= 68

0.608

0.458***
Not
Significant

212

6.4.7

Conclusions from Experiment 2


This experiment examined the effects of varying threat levels and

goal frames on PMT variables and the subsequent effects of the PMT
variables on message involvement, attitudes and purchase intention. The
manipulation checks were moderately successful as under higher threat
conditions, participants identified the threat levels correctly. Internal
consistency of the self-efficacy scale successfully improved to 0.56. There
was no main effect of threat levels on perceived severity of threat, perceived
vulnerability and fear. The goal frames did not influence the threat appraisal
variables as hypothesized. There was also no significant interaction effect
between the factors as hypothesized. However independent variables had an
interaction effect on perceived vulnerability. The interaction showed that gain
frames and higher threat levels increased feelings of perceived vulnerability.
This result confirms the results of previous studies (Rothman et al 1993;
Mann et al 2004) that gain frames work better in the case of preventive
behaviour. This result also shows that goal framing can be used to promote
pro-environmental behaviour by accentuating intrinsic goals related to health
and well-being (Lindenberg & Steg 2007; Pelletier & Sharp 2008).
In contrast to previous studies (Cox & Cox 2001; Meyers-Levy &
Maheswaran 2004; van t Riet et al 2008;OKeefe & Jensen 2009; Janssens
et al 2010; Updegraff 2013) loss frames did not increase threat perception.
Similarly there was no relationship between involvement and framing. Both
environmental concern and message involvement did not interact with frames
to produce an effect on purchase intention.
Perceived severity and perceived vulnerability significantly
predicted fear as proposed by PMT (Rogers & Prentice-Dunn 1997; Floyd et
al 2000) and other studies that apply this theory (Milne et al 2000; de Hoog et
al 2008). Since the participants judged the threat to be high, fear levels

213

increased and coping appraisal was also intiated (Maddux & Rogers 1983;
Boer & Seydel 1996; Milne et al 2000). In the case of the previous
experiment only response efficacy and perceived vulnerability predicted
message involvement. In this experiment all the PMT variables (perceived
severity, perceived vulnerability, fear and response efficacy) except self
efficacy significantly predicted message involvement. This shows that higher
levels of health risk increased elaboration and thereby increased their
involvement with the message as observed in previous studies (Bloch &
Richins 1983; Richins & Bloch 1986; Keller & Block 1996; de Hoog 2005).
Therefore fear can increase more effortful processing in environment related
communication (Meijnders et al 2001). The relationship between the PMT
variables and message involvement is similar to the results presented by
recent research (Cauberghe et al 2009). Environmental knowledge did not
have any effect on the hypothesized variables. Unlike the previous
experiment, environmental concern had a significant effect on message
involvement, attitudes and intentions. This confirms ELMs proposition that
issue involvement has an effect on attitudes and intentions (Petty & Cacioppo
1986). Both environmental concern and message involvement had a
significant influence on attitudes and intention similar to other advertising
studies (Gardner 1985; Park & Young 1986; Muehling & Laczniak 1988).
The manipulation checks were comparatively successful than
Experiment1 for the threat level perception. Therefore Experiment 3 was
designed using these two factors (threat level and goal frames) by refining the
stimuli. Since high scores for perceived severity and vulnerability were
reported in both the scenarios, the threat levels were modified, such that the
low threat level contained very generic statements about pollution issues,
whereas the high threat levels highlighted the perceived vulnerability to the
threat, by mentioning risks of cancer and respiratory illness. The next

214

experiment used the issue of e-waste (based on mobile phone stimuli) to


observe if similar effects were observed.
6.5

STUDY 2: EXPERIMENT 3: THREAT LEVELS AND GOAL


FRAMING (MOBILE PHONE STIMULI)
Experiment 3 was designed to use threat levels and goal frames to

influence the PMT variables. This experiment was conducted with mobile
phones as the chosen product. The mobile phone stimulus was chosen for the
experiment and the changes to the stimulus were made based on the
information gained from Experiment 2.
6.5.1

Experimental Design
A 2 (frame type: gain vs. loss) x 2 (threat level: high vs. low)

between subjects experimental design was utilized to investigate the


hypotheses. This resulted in four possible combinations of the stimuli. One
hundred and ninety undergraduate engineering students from a large South
Indian University (52.4 % male, median age=20) were randomly assigned to
the four possible conditions. Males and female respondents were represented
almost equally. The respondents age ranged from 18-22. The participants
were exposed to the stimulus and data collection was through a paper and
pencil questionnaire.The questionnaire (Q4) is shown in Appendix 7. On
completing the questions on the dependent variables, the respondents were
given the filler questionnaire followed by the counterbalanced questionnaires
related to the environment (environmental concern and environmental
knowledge). This was similar to the questions in Experiment 1. The
personality variable CFC was also not included in this questionnaire.

215

6.5.2

Stimuli and Treatment Validity


A total of four print advertisements were developed for the four

cells: high threat level and gain frame; high threat level and loss frame; low
threat level and gain frame; low threat level and loss frame.

The

advertisements listed the features of the green mobile phone and specified that
the mobile phone is 82% recyclable. In Experiment 2, the participants judged
low levels of threat as a severe threat. Hence in this experiment the threat
levels were toned down in the low threat level conditions.
In the low threat condition, the advertisement emphasized that ewaste was difficult to dispose and did not specifically mention a health threat.
In the high threat condition, the message emphasized health threats like
respiratory illness and highlighted personal vulnerability towards the threat.
The gain frame exhorted the respondents to protect themselves and the loss
frame message highlighted the potential losses incurred when not purchasing
a green product. These were also made stronger. The advertisements were
shown to the panel as in the previous experiments to check their validity. The
advertisements are presented in Appendix 8 (Figure A8.1, Figure A8.2, Figure
A8.3 and Figure A8.4).
6.5.3

Manipulation Checks
Threat level manipulations were checked by verifying if the

perceived severity and perceived vulnerability varied for different threat


levels. Frame manipulations were checked by asking the respondents (similar
to Experiment 2) to rate the following questions: I can gain health benefits
by buying recylable products, I can lose important health benefits if I dont
buy recylable products. The response to this items was measured using seven
point Likert scales anchored from 1 = Strongly Disagree and 7 = Strongly
Agree.

216

6.5.4

Dependent Variables
Most of the dependent variables were the same as Experiment 2.

The items were changed to reflect the issue of e-waste and changes are
described below.
6.5.4.1

Protection motivation theory variables

Perceived severity
Perceived severity was measured using a three item seven point
scale where 1 = Strongly Disagree and 7 = Strongly Agree. Participants were
asked to indicate their responses on the following statements. I believe that
e-waste pollution is a serious threat to human health, I believe that e-waste
disposal may cause severe health issues, I believe that e-waste pollution is
extremely harmful. The three items were averaged into a single perceived
severity score.
Perceived vulnerability
Perceived vulnerability was measured using a three item seven
point scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine
the participants susceptibility to the threat. Participants were asked to
indicate their responses on the following statements: It is possible that I
might get affected by diseases caused by e-waste pollution., It is probable
that I will suffer from various diseases caused by e-waste pollution, I am at
risk for getting health problems caused by e-waste pollution. These items
were collapsed into a single perceived vulnerability score.

217

Response efficacy
Response efficacy was also measured using a three item seven
point scale where 1 = Strongly Disagree and 7 = Strongly Agree to determine
whether participants believed if purchasing recyclable products averted the
threat. Participants rated their responses on the following statements:
Purchasing recyclable products is a highly effective way of preventing
diseases caused by e-waste pollution, Buying recyclable products will
significantly lower my risk of being affected by diseases caused by e-waste
pollution, Buying recyclable products is an effective method of reducing
threats to human health caused by e-waste pollution These items were
combined into a single response efficacy score.
Self efficacy
Self efficacy was measured using a three item seven point
scaleswhere 1 = Strongly Disagree and 7 = Strongly Agree to determine
whether participants believed if they were capable of averting the threat.
Participants rated their responses on the following statements: I am capable
of checking if products contain recyclable materials, I can easily switch
over to recyclable products to prevent future health problems, I can identify
and purchase recyclable products These items were combined into a single
self efficacy score. The rest of the dependent variables remained the same as
Experiment 2. The questionnaire (Q4) is shown in Appendix 7 as discussed
previously.
6.5.5

Demographics
There were very few demographic variables collected from the

participants, as these variables were not the major focus of the research
questions. Table 6.60 shows the demographic details of the sample based on

218

the usable responses out of the 190 students. Male and female students were
represented in almost equal proportions. Most of the students were aged 20
and above.
6.5.6

Data Screening
Unlike previous experiments, this study planned to use PLS-SEM.

Hence to ensure data integrity, data screening was conducted using SPSS 19
to ensure the validity of the data prior to hypotheses testing. Data was first
screened for missing values and outliers (univariate and multivariate). Apart
from this, other multivariate statistical assumptions (normality, linearity and
homoscedasticity) were also investigated.
Table 6.60 Demographic details
Variable

Count

Gender
Male

96

47

Female

85

53

Total

181

100

18

0.6

19

35

19.3

20

96

53.0

21

44

24.3

22

2.8

42

23.2

Electrical and Electronics Engineering 56

30.9

Information Technology

31

17.1

Mining Engineering

15

8.3

Printing Technology

37

20.4

Age

Course
Civil Engineering

219

6.5.6.1

Missing Value Analysis


On inspection, there were very few missing values. Only one value

was missing in the dependent variable. Hair et al (2009) recommend mean


imputation as a suitable method for replacing missing values in such cases.
Hence mean imputation was used to replace this missing value. After
imputation of the missing value, the items were summated to determine the
composite score of the variables.
6.5.6.2

Outlier Analysis

Univariate outliers
Box plots were used to identify univariate outliers among
dependent variables. Outlier analyses were conducted on composite variables
to reduce the effect any variations that single indicators might cause. Few
outliers were identified.
Multivariate outliers
Multivariate outliers are unusual combinations of variable values
(Hair et al 2009). Mahalanobis' distance was used to identify multivariate
outliers. A conservative level of significance of 0.001 was used to identify
outliers. Four cases were identified as outliers.
6.5.6.3

Decision regarding outliers


Since there were no data entry errors or other anomalies, univariate

outliers were not deleted as deleting them might impact the generalizability of
the data (Hair et al 2009). Only the multivariate outliers were removed as they
represented a small proportion (2.1%) of the total number of cases. This
resulted in 181 usable responses.

220

6.5.6.4

Univariate normality
Skewness and kurtosis were assessed to determine univariate

normality. Certain variables (perceived severity, response efficacy, self


efficacy, message involvement) did not follow a normal distribution.
For instance, respondents perceived high severity in both high and low threat
conditions and therefore the perceived severity value was negatively skewed.
The skewness and kurtosis values are shown in Table 6.61.
Table 6.61 Univariate normality
Mean

Skewness

Statistic Statistic

Kurtosis

Zskew

Statistic

Zkurtosis

PERC_SEV

5.6427

-.900

-4.97238

.696

1.938719

PERC_VUL

4.7403

-.681

-3.76243

.402

1.119777

RESP_EFFICACY

5.6446

-.816

-4.50829

1.287

3.584958

SELF_EFFICACY

4.6961

-.330

-1.8232

-.434

-1.20891

MESSAGE_INVOLVEMENT 5.1860

-.766

-4.23204

.755

2.103064

FEAR

3.9945

-.221

-1.22099

-.563

-1.56825

ATTITUDE_AD

5.3168

-.632

-3.49171

.500

1.392758

ATTITUDE_BRAND

5.1731

-.321

-1.77348

-.341

-0.94986

PURCHASE_INTENTION

4.5783

-.544

-3.00552

-.143

-0.39833

ENV_KNOW

6.6133

.360

1.98895

-.076

-0.2117

TOTAL_ENV_CONCERN

5.9982

-1.420

-7.8453

3.451

9.612813

Most of the skewness and kurtosis values range from -1 to +1,


except environmental concern. Further tests showed that Z-values of the
skewness and kurtosis values of most variables were negative. This is not
surprising as the computed mean values are very high for the threat and fear
variables. Hair et al (2009) also recommend Z-tests for testing the skewness

221

of the variables and based on significance at 0.01 levels, seven variables


exceeded the value of 2.58. Most variables are negatively skewed.
6.5.6.5

Multivariate Statistical Assumptions

Multivariate normality
Multivariate normality was assessed based on Mardias coefficient
(Mardia 1970) using IBM AMOS 18. A high critical ratio of the coefficient
(26.512) indicated that data was significantly not normal as it exceeded the
cut-off value of 5.0 as suggested by Bentler (2006).
Linearity and homoscedasticity
Linearity and homoscedasticity were assessed among the variables
by using the regression residual and scatter plots. The variables met the
assumptions of linearity and homoscedasticity.
6.5.6.6

Conclusions from Data Screening


There was very few missing data and data imputation was done to

handle missing data. Data analysis showed that most variables were
negatively skewed and had a non-normal distribution. MANOVA is robust to
the violations of normality (Leech et al 2011) and therefore hypotheses
involving MANOVA were conducted using IBM SPSS 19. This study was
planned to be analysed using partial least squares based SEM (PLS-SEM) and
therefore this choice ensured that non normal data distribution did not pose a
problem for further analysis. Additionally the fact that PLS also works well
for a series of cause and effect relationships is to the studys advantage
(Bontis et al 2007). PLS bootstrap also provides a more accurate and efficient
estimation of structural model parameters when compared to MLE and

222

Bollen-Stine SEM bootstraps when there are fewer than 200 observations
(Sharma & Kim 2013).
6.5.7

Validity and reliability analyses


Validity and reliability was assessed in two different ways. For the

first level of analysis, to assess the effect of the factors on the PMT variables
MANOVA is to be used. Hence Cronbachs

& EFA (Exploratory Factor

Analysis) were used to analyse the reliability and validity of the PMT
variables.
Next, the results of path analysis were to be analysed using PLSSEM. Hence, the measurement model was checked to ensure the reliability
and validity criteria associated with the formative and reflective measurement
model.
6.5.7.1

Validity and reliability of the PMT variables


Exploratory Factor Analysis (EFA) was used to examine

component loadings for the PMT constructs. The prescribed minimum sample
size for EFA is 100 (Hair et al 2009). On completion of the EFA, scale
reliabilities were assessed using the reliability coefficient (Cronbach ). The
PMT constructs (perceived severity, perceived vulnerability, fear, response
efficacy and self efficacy) loaded on five factors. Since PMT hypothesizes
close relationship between the constructs, oblique rotation was used and
factor loading above 0.4 was used to interpret each factor (Wu et al 2005).
None of the items loaded on more than one factor.

Next, the internal

reliabilities of the constructs were tested using Cronbachs


adequate.

and were found

223

Scale reliability of the PMT constructs


Table 6.62 lists the reliabilities of the PMT constructs used. All the
scales had adequate reliabilities except response efficacy which had a
moderate reliability of 0.56. However, other PMT studies have also used this
variable despite achieving such moderate reliability scores (e.g. Milne et al
2002; Wu et al 2005; Daley et al 2009). Hence, this variable was retained in
the study.
Table 6.62 Experiment 3: Means and reliabilities of the PMT constructs
Scale Items

Item-total correlation ( ) MEAN

SD

Perceived Severity

0.67

5.65

0.93

Perceived Vulnerability

0.62

4.76

1.13

Response Efficacy

0.56

5.64

0.82

Self Efficacy

0.61

4.7

1.18

Fear

0.87

4.0

1.321

6.5.7.2

Reliability of the reflective constructs by assessing the


measurement/outer model
Indicator reliability and composite reliability were used to assess

the reliability of the reflective measurement model. Cronbachs

exceeded

0.6 for all constructs (Table 6.62), except response efficacy which had a
moderate reliability as discussed previously. Composite reliability was now
used to prioritise indicators during estimations. Indicator reliability was
checked by examining the indicator loadings. Table 6.63 shows the indicator
loading and the composite reliability. The loadings ranged from 0.40 to 0.94
and most of them exceeded 0.70 (Fornell & Larcker 1981). Loadings below
0.7 are candidates for deletion if deleting these indicators leads to an increase
in composite reliability above the threshold value of 0.70. However composite

224

reliability scores ranged from 0.76 to 0.95 and therefore exceeded the 0.7 cutoff (Hair et al 2011; Hair et al 2012). Only MI2 (The ad's message seemed
relevant to me.) from the message involvement scale had a low value but it
was not necessary to remove this indicator as the AVE and CR values were
above requisite cut-off criteria (Hair et al 2011) .
Common method bias
Common method variance is a potential problem in social science
research as data is collected through surveys based on self reporting. This
research tried to minimise the effect of common method bias by (a) increasing
the clarity of questions by using iterative pretests (b) not collecting sensitive
personal data as it might induce social desirability responses (Herath & Rao
2009; Mohan et al 2013). Harmans One-Factor Test was employed to test if
a single factor emerged from the analysis or if a general factor explained
majority of the variance (Podsakoff et al 2003). The results of an exploratory
factor analysis on IBM SPSS 19 showed that multiple factors were present
and the major factor accounted for only 29 % of the total variance. However
this method has its own limitations (Podsakoff et al 2003; Chin et al 2012).
Therefore the correlations matrix of the latent variables was observed and the
largest correlation was 0.65 which is lower than the correlations that suggest
common method bias (r > 0.9).
6.5.7.3

Validity
Convergent and discriminant validity was also assessed for the

measurement model. The average variance extracted (AVE) was higher than
the requisite 0.50 for all the constructs except environmental concern which is
a second order construct. The values are shown in Table 6.63. Discriminant
validity was evaluated by the examination of the cross loading of the variable
and the Forner-Larcker criterion.

225

Table 6.63

Experiment 3: Composite reliability, indicator reliability


and convergent validity

Construct
Perceived Severity

Perceived Vulnerability

Response Efficacy

Self Efficacy

Fear

Message Involvement

Attitude towards brand

Attitude towards ad

Outer
Loading
>0.7

Composite
Reliability
>0.7

AVE

PS1

0.8361

0.824

0.619

PS2
PS3

0.8671
0.6251

PV1
PV2

0.8741
0.5212

0.791

0.568

PV3
RE1
RE2
RE3

0.8177
0.7463
0.7732
0.6812

0.778

0.539

SE1

0.6773

0.761

0.525

SE2
SE3

0.5718
0.8864

F1
F2
F3
F4

0.8286
0.8871
0.8813
0.7723

0.904

0.661

F5
MI1
MI2

0.6578
0.6895

0.859

0.512

0.951

0.866

0.901

0.753

Indicator

MI3
MI4
MI5
MI6

0.4081
0.7837
0.7461
0.7964
0.7968

AAB1
AAB2

0.917
0.9402

AAB3

0.935

AAD1

0.8815

AAD2

0.8629

AAD3

0.8559

>0.5

226

Table 6.63 (Continued)

Construct
Purchase Intention

Biospherical-Concern

Egoistic-concern

Altruistic - Others

Environmental concern

Outer
Loading
>0.7

Composite
Reliability
>0.7

AVE

PI1

0.9222

0.949

0.860

PI2

0.9301

PI3

0.933

Animals

0.711

0.844

0.573

Birds

0.778

Plants

0.803

Children

0.741

Me

0.833

0.844

0.538

My Future

0.839

My Health

0.708

My LifeStyle

0.643

All People

0.715

0.805

0.626

My children

0.684

People in
my country

0.714

Marine

0.736
0.859

0.604

Indicator

Bio

>0.5

Egoistic
Altruistic

The AVE of the latent construct must be greater than the latent
constructs highest squared correlation with other constructs (Fornell &
Larcker 1981). It can be seen from Table 6.64 that the Forner-Larcker
criterion is satisfied. Table 6.65 shows the details of the second-order
construct. Here too, the criterion was met.
The main diagonal in Table 6.64 and 6.65 show the AVE of the
constructs. The scales satisfied the discriminant validity criteria. The loadings
and cross-loading of item to other constructs were also inspected to evaluate

227

discriminant validity. Items loaded more on their constructs when compared


to other constructs as required.
Table 6.64 Experiment 3: Discriminant validity
AAD AAB

FEAR MI

PS

PV

PI

RE

SE

AAD 0.753
AAB 0.498 0.866
FEAR 0.071 0.068 0.661
MI

0.313 0.281 0.26

0.512

PS

0.0595 0.091 0.148

0.127 0.619

PV

0.053 0.094 0.086

0.089 0.164 0.568

PI

0.243 0.428 0.132

0.231 0.075 0.034 0.86

RE

0.088 0.071 0.094

0.181 0.261 0.158 0.047961 0.539

SE

0.033 0.035 0.024

0.071 0.076 0.031 0.09

Table 6.65

0.127449 0.525

Experiment 3: Discriminant validity of the second order


construct (environmental concern)
Egoistic

Altruistic

Egoistic

0.538

Altruistic

0.099

0.573

Biospheric

0.130

0.298

Biospheric

0.626

The analysis shows that the reflective measurement model for the
(both first-order and second-order) variables used in this research are reliable
and valid.
Validity of the formative construct
Hair et al (2011) recommend the examination of convergent
validity, collinearity among indicators and use previous theory to retain
indicators that do not have significant outer loading to asses the formative

228

constructs and indicators. The weight of the indicators was examined by


resampling using bootstrapping (181 observations per subsample, 5,000
subsamples and no sign changes) in SmartPLS as this is the primary statistic
for examining the indicators (Hair et al 2012). The t-values were significant
for most of the indicators except EK7, EK8, EK14 and EK15 (p<0.05). Hence
these indicators were removed from the model. Subsequently variance
inflation factor (VIF) was used to test the multicollinearity among the
remaining environment knowledge indicators. The results show minimal
collinearity among the indicators as the VIF of all items ranged between
1.073 and 1.33, below the common cut off

value of 5. Therefore, the

assumption of multicollinearity was not violated (Chin, 2010).


6.5.8

Results of Experiment 3: threat levels and goal framing


This experiment was conducted to verify goal framing and threat

levels. The mobile phone stimulus was chosen, since electronic waste seemed
to be a less familiar issue when compared to biodegradability.
This study was conducted with the mobile phone stimuli to evaluate
the effect of different threat levels (low/high) and goal frames (gain vs. loss)
on the PMT variables. The effect of PMT variables on involvement and the
subsequent influence of involvement on attitudes and purchase intention were
evaluated using the path model.
6.5.8.1

Manipulation checks
One-way ANOVA was conducted to examine the effectiveness of

the message in manipulating the perceived severity, perceived vulnerability


based on threat levels. In both the threat levels, the mean for the perceived
severity remained above 5.5 and perceived vulnerability scores ranged above
4.9. Therefore there was no statistically significant effect of the threat level on

229

these variables. Similarly the manipulation checks of the frames indicated that
there was no statistically significant effect of the frames on the variables
included for manipulation checks. Tables 6.66a, 6.66b, 6.66c and 6.66d show
the results of the manipulation check.
Table 6.66a Experiment 3: Manipulation check: effect of threat level on
perceived severity
Descriptive Statistics
Dependent Variable:PERCEIVED SEVERITY
Threat Level

Mean

Std. Deviation

high

5.7753

.88731

89

low

5.5145

.96383

92

Total

5.6427

.93362

181

Tests of Between-Subjects Effects


Dependent Variable:PERCEIVED SEVERITY
Source

Type III Sum of Squares df Mean Square

Corrected Model 3.077

Sig.
.060

3.077

3.580

Intercept

5765.935

5765.935

6709.822 .000

Threat_level

3.077

3.077

3.580

Error

153.820

179 .859

Total

5920.000

181

Corrected Total 156.896

.060

180

a. R Squared = .020 (Adjusted R Squared = .014)

Table 6.66b Experiment 3: Manipulation check: effect of threat level on


perceived vulnerability
Descriptive Statistics
Dependent Variable:PERCEIVED VUNERABILITY
Threat Level

Mean

Std. Deviation

High

4.8052

1.02840

89

low

4.6775

1.24177

92

Total

4.7403

1.14049

181

230

Table 6.66b (Continued)


Tests of Between-Subjects Effects
Dependent Variable:PERCEIVED VULNERABILITY
Source

Type III Sum of Squares df Mean Square

Sig.

Corrected Model .738a

.738

.566

Intercept

4067.903

4067.903

3119.890 .000

Threat_level

.738

.738

.566

Error

233.391

179 1.304

Total

4301.333

181

Corrected Total 234.129

.453
.453

180

a. R Squared = .003 (Adjusted R Squared = -.002)

Manipulation checks also revealed that the frame manipulation was


not successful as there was no significant main effect of the frame
manipulation (loss vs. gain) in both the conditions (Table 6.66c and 6.66d).
Table 6.66c Experiment 3: Manipulation check: effect of frame type
(gain)
Tests of Between-Subjects Effects
Dependent Variable:MC_GAIN
Source

Type III Sum of Squares df Mean Square

Sig.

Corrected Model .816a

.816

.526

Intercept

5045.855

5045.855

3251.132 .000

frame_type

.816

.816

.526

Error

277.813

179 1.552

Total

5328.000

181

Corrected Total 278.630

180

a. R Squared = .003 (Adjusted R Squared = -.003)

.469

.469

231

Table 6.66d Manipulation check: effect of frame type (loss)


Tests of Between-Subjects Effects
Dependent Variable:MC_LOSS
Source

Type III Sum of Squares df Mean Square

Corrected Model

.021

.021

3916.154

3916.154

.021

.021

Error

461.051

179

2.576

Total

4378.000

181

Corrected Total

461.072

180

Intercept
frame_type

Sig.

.008

.928

1520.422 .000
.008

.928

a. R Squared = .000 (Adjusted R Squared = -.006)

Similar to the other experiments , failed manipulation checks were


not of great concern and therefore further analysis on the data was conducted.
6.5.8.2

Hypotheses tests of the effect of manipulations on PMT


variables
The hypothesized effect of goal frames and threat levels on the

PMT variables was analyzed using MANOVA. The dependent variables


(perceived severity, perceived vulnerability, fear, response-efficacy and selfefficacy) were only moderately correlated (0.21 0.51) and therefore there
was no risk of multicollinearity to pose a hindrance to MANOVA. Tables
6.67 and 6.68 show the distribution characteristics and the group wise means
of the protection motivation variables. Similar to previous experiments, the
perceived severity and perceived vulnerability to the threat are on the higher
side. The group wise means do not seem to differ much similar to the
previous experiments.

232

Table 6.67

Experiment 3: Distribution characteristics of the protection


motivation variables watch stimuli
Minimum Maximum Mean Std. Deviation

PERC_SEV

2.67

7.00

5.64

0.93

PERC_VUL

1.00

7.00

4.74

1.14

RESP_EFFICACY

2.33

7.00

5.64

0.83

SELF_EFFICACY

1.67

7.00

4.70

1.20

FEAR

1.00

7.00

3.99

1.33

Table 6.68

Experiment 3: Group wise mean values of protection


motivation variables for the mobile phone stimuli
Self

Factor

Perceived
Severity

Perceived
Vulnerability

Response
Efficacy

Efficacy

Threat
level: High

5.77

4.80

5.71

4.78

3.98

Threat
level: Low

5.51

4.67

5.57

4.60

4.00

Goal frame:
Gain

5.72

4.72

5.71

4.66

3.86

Goal frame:
Loss

5.56

4.75

5.57

4.72

4.12

Fear

A one-way MANOVA was conducted to test hypothesis 5 (H5) that


stated that participants who viewed advertisements with higher threats levels
would report higher levels of severity and vulnerability when compared to
consumers who viewed weaker threats. The results did not show significant
differences between the groups (Pillais Trace=0.20; Wilks lambda = 0.980;
Hotellings Trace and Roys Largest Root = 0.020, F(2,178) =1.782, p >0.05)
and hence the hypothesis was not supported (Table 6.69a and 6.69b).

233

Table 6.69a Experiment 3: Hypothesis 5: multivariate tests (mobile


phone stimuli)
Multivariate Testsb
Effect
Intercept

Value

Hypothesis
df

Error
df

Sig.

Pillai's Trace

.977 3712.945a

2.000

178.000 .000

Wilks' Lambda

.023 3712.945a

2.000

178.000 .000

Hotelling's
Trace

41.718 3712.945

2.000

178.000 .000

Roy's Largest
Root

41.718 3712.945a

2.000

178.000 .000

.020

1.782a

2.000

178.000 .171

.980

1.782a

2.000

178.000 .171

Hotelling's
Trace

.020

1.782

2.000

178.000 .171

Roy's Largest
Root

.020

1.782a

2.000

178.000 .171

Threat_level Pillai's Trace


Wilks' Lambda

a. Exact statistic b. Design: Intercept + Threat_level

Table 6.69b Experiment 3: Hypothesis 5: tests of between-subjects


effects (mobile phone stimuli)
Tests of Between-Subjects Effects
Source
Corrected
Model

Intercept

Dependent
Variable

Type III Sum of


df
Squares

Mean
Square

Sig.

Perceived
severity

3.077 a

3.077

3.580

.060

Perceived
vunerability

.738b

.738

.566

.453

Perceived
severity

5765.935

5765.935 6709.822 .000

Perceived
vunerability

4067.903

4067.903 3119.890 .000

234

Table 6.69b (Continued)


Tests of Between-Subjects Effects
Dependent
Variable

Source
Threat_level

Error

Total

Corrected
Total

Type III Sum of


df
Squares

Mean
Square

Sig.

Perceived
severity

3.077

3.077

3.580

.060

Perceived
vunerability

.738

.738

.566

.453

Perceived
severity

153.820

179

.859

Perceived
vunerability

233.391

179

1.304

Perceived
severity

5920.000

181

Perceived
vunerability

4301.333

181

Perceived
severity

156.896

180

Perceived
vunerability

234.129

180

a. R Squared = .020 (Adjusted R Squared = .014)


b. R Squared = .003 (Adjusted R Squared = -.002)

Hypothesis 6 (H6) was not supported as the results indicated that


there was no statistically significant difference in severity and vulnerability
based on frame type (Pillais Trace=0.010; Wilks lambda = 0.990;
Hotellings Trace and Roys Largest Root = 0.010, F(2,178) = 0.872, p >0.05)
(Table 6.70a and 6.70b).
perception.

Therefore frame type did not increase threat

235

Table 6.70a Experiment 3: Hypothesis 6:


phone stimuli)

multivariate tests (mobile

Multivariate Testsb
Value

Hypothesis
df

Error
df

Sig.

Pillai's Trace

.976

3659.959a

2.000

178.000

.000

Wilks' Lambda

.024

3659.959 a

2.000

178.000

.000

Hotelling's
Trace

41.123 3659.959a

2.000

178.000

.000

Roy's Largest
Root

41.123 3659.959a

2.000

178.000

.000

Effect
Intercept

.010

.872a

2.000

178.000

.420

Wilks' Lambda

.990

.872a

2.000

178.000

.420

Hotelling's
Trace

.010

.872a

2.000

178.000

.420

Roy's Largest
Root

.010

.872a

2.000

178.000

.420

frame_type Pillai's Trace

a. Exact statistic
b. Design: Intercept + frame_type

Table 6.70b Experiment 3: Hypothesis 6: tests of between-subjects


effects (mobile phone stimuli)
Tests of Between-Subjects Effects
Source
Corrected
Model

Intercept

Dependent
Variable

Type III Sum


of Squares

df

Mean
Square

Sig.

Perceived
severity

1.124 a

1.124

1.292

.257

Perceived
vunerability

.054b

.054

.041

.840

Perceived
severity

5764.188

5764.188

6623.707 .000

Perceived
vunerability

4065.598

4065.598

3109.007 .000

236

Table 6.70b (Continued)


Tests of Between-Subjects Effects
Dependent
Variable

Source
frame_type

Error

Total

Corrected
Total

Type III Sum


of Squares

df

Mean
Square

Sig.

Perceived
severity

1.124

1.124

1.292

.257

Perceived
vunerability

.054

.054

.041

.840

Perceived
severity

155.772

179

.870

Perceived
vunerability

234.075

179

1.308

Perceived
severity

5920.000

181

Perceived
vunerability

4301.333

181

Perceived
severity

156.896

180

Perceived
vunerability

234.129

180

a. R Squared = .007 (Adjusted R Squared = .002)


b. R Squared = .000 (Adjusted R Squared = -.005)

The proposed interaction between threat levels and frames was also
not supported (Pillais Trace=0.011; Wilks lambda = 0.989; Hotellings
Trace and Roys Largest Root = 0.011, F(3,175) = 0.630, p >0.05).
Therefore hypothesis 7 (H7) was not supported. Tables 6.71a and 6.71b show
the MANOVA results. Therefore the factors did not interact to produce any
significant results.

237

Table 6.71a Experiment 3: Hypothesis 7:


phone stimuli)

multivariate tests (mobile

Multivariate Testsb
Effect
Intercept

frame_type

Threat_level

frame_type *
Threat_level

Value

Hypothesis
df

Error
Sig.
df

Pillai's Trace

.977 3720.629a

2.000

176.000 .000

Wilks' Lambda

.023 3720.629a

2.000

176.000 .000

Hotelling's
Trace

42.280 3720.629a

2.000

176.000 .000

Roy's Largest
Root

42.280 3720.629a

2.000

176.000 .000

Pillai's Trace

.010

.876a

2.000

176.000 .418

Wilks' Lambda

.990

.876a

2.000

176.000 .418

Hotelling's
Trace

.010

.876a

2.000

176.000 .418

Roy's Largest
Root

.010

.876a

2.000

176.000 .418

Pillai's Trace

.020

1.810a

2.000

176.000 .167

Wilks' Lambda

.980

1.810

2.000

176.000 .167

Hotelling's
Trace

.021

1.810a

2.000

176.000 .167

Roy's Largest
Root

.021

1.810a

2.000

176.000 .167

Pillai's Trace

.010

.901a

2.000

176.000 .408

Wilks' Lambda

.990

.901

2.000

176.000 .408

Hotelling's
Trace

.010

.901a

2.000

176.000 .408

Roy's Largest
Root

.010

.901a

2.000

176.000 .408

a. Exact statistic
b. Design: Intercept + frame_type + Threat_level + frame_type * Threat_level

238

Table 6.71b Experiment 3: Hypothesis 7: tests of between-subjects


effects (mobile phone stimuli)
Tests of Between-Subjects Effects
Source
Corrected
Model

Dependent
Variable

Type III
Sum of
Squares

df

Perceived
5.210 a
3
severity
Perceived
2.467 b
3
vunerability
Intercept
Perceived
5766.088
1
severity
Perceived
4065.379
1
vunerability
frame_type
Perceived
1.140
1
severity
Perceived
.046
1
vunerability
Threat_level
Perceived
3.115
1
severity
Perceived
.777
1
vunerability
frame_type *
Perceived
1.029
1
Threat_level
severity
Perceived
1.673
1
vunerability
Error
Perceived
151.686
177
severity
Perceived
231.662
177
vunerability
Total
Perceived
5920.000
181
severity
Perceived
4301.333
181
vunerability
Corrected Total Perceived
156.896
180
severity
Perceived
234.129
180
vunerability
a. R Squared = .033 (Adjusted R Squared = .017)
b. R Squared = .011 (Adjusted R Squared = -.006)

Mean
Square

Sig.

1.737

2.027

.112

.822

.628

.598

5766.088

6728.366 .000

4065.379

3106.127 .000

1.140

1.330

.250

.046

.035

.851

3.115

3.635

.058

.777

.593

.442

1.029

1.201

.275

1.673

1.278

.260

.857
1.309

239

A regression analysis was done with three predictors: framing,


environmental concern and the interaction term with purchase intention as the
dependent variable to test H8a. Framing was dummy coded with the lossframe message condition allocated a value of 0 and the gain-frame message
condition a value of 1. The interaction terms were calculated as a product of
frame type and environmental concern (frame x environmental concern) from
these variables. It can be seen from Table 6.72 that the hypothesis was not
supported as interaction between the variables did not predict purchase
intention. Hence the two variables did not have the hypothesized effect.
Table 6.72a Experiment 3: Hypothesis 8a: interaction of frame and
environmental concern on purchase intention (mobile phone
stimuli)
Model Summary
Model

R Square

Adjusted R Square

Std. Error of the


Estimate

.0.127a

0.016

-0.001

1.45332

a. Predictors: (Constant), TOTAL_ENV_CONCERN, FRAME_CODED,


ENV_CONC_X_FRAME
ANOVAb
Sum of
Squares

df

Mean
Square

Sig.

6.123

2.041

.966

.410a

Residual

373.796

177

2.112

Total

379.919

180

Model
1

Regression

a. Predictors: (Constant), TOTAL_ENV_CONCERN, frame_coded,


frame_x_env_concern
b. Dependent Variable: PURCHASE_INTENTION

240

Table 6.72a (Continued)


Coefficientsa
Unstandardized
Coefficients

Model

Standardized
Coefficients

Sig.

Std.
Error

5.118

2.463

-1.237

1.682

-.427

-.735 .463

frame_x_env_concern

.186

.278

.413

.668

TOTAL_ENV_CONCERN

-.059

.407

-.033

-.146 .884

1 (Constant)
frame_coded

Beta
2.078 .039
.505

a. Dependent Variable: PURCHASE_INTENTION

Similarly Table 6.72b shows that H8b was not supported as


message involvement did not interact with frame type to produce an effect on
purchase intentions. However, the model was significant, a follow up
stepwise regression revealed that only message involvement significantly
predicted purchase intention.
Table 6.72b Experiment 3: Hypothesis 8b: interaction of frame and
message involvement on purchase intention (mobile phone
stimuli)
Model Summary
Model R
R Square Adjusted R Square
a
1
.0.474
0.225
0.212

Std. Error of the Estimate


1.28982

Predictors: (Constant), frame_x_MI, MESSAGE_INVOLVEMENT, frame_coded


ANOVAb
Model
1

Regression

Sum of
Squares
85.455

df
3

Mean
Square
28.485

Sig.

17.122

.000a

Residual
294.464
177
1.664
Total
379.919
180
a. Predictors: (Constant), frame_x_MI, MESSAGE_INVOLVEMENT,
frame_coded b. Dependent Variable: PURCHASE_INTENTION

241

Table 6.72b (Continued)


Coefficientsa
Unstandardized Standardized
Coefficients
Coefficients

Model

B
1 (Constant)
.830
frame_coded
.043
MESSAGE_INVOLVEMENT .768
frame_x_MI

-.038

Sig.

Std.
Error
1.618
1.060
.307

.015
.509

.513 .609
.040 .968
2.500 .013

.201

-.079

-.189

Beta

.850

a. Dependent Variable: PURCHASE_INTENTION

A one-way MANOVA analysis of the factors and gender showed


significant differences in risk perception based on gender. Therefore H8c was
supported. Tables 6.72c and 6.72d show the results. It can be seen that gender
has an effect on perceived severity and fear independently ((Pillais
Trace=0.058; Wilks lambda = 0.942; Hotellings Trace and Roys Largest
Root = 0.061, F(3,171) = 3.495, p <0.05). Gender also interacts with the
factors to produce an effect on the perceived severity of threat (Pillais
Trace=0.047; Wilks lambda = 0.953; Hotellings Trace and Roys Largest
Root = 0.050, F(3,171) = 2.834, p <0.05).
Table 6.72c Experiment 3: Hypothesis 8c: multivariate tests (mobile
phone stimuli)
Multivariate Tests b
Effect
Intercept

Hypothesis
Error df Sig.
df

Value

Pillai's Trace

.978

2572.816a

3.000

171.000 .000

Wilks'
Lambda

.022

2572.816a

3.000

171.000 .000

45.137 2572.816a

3.000

171.000 .000

Roy's Largest
45.137 2572.816a
Root

3.000

171.000 .000

Hotelling's
Trace

242

Table 6.72c (Continued)


Multivariate Tests b
Effect
frame_type

Threat_level

GENDER

frame_type *
Threat_level

Hypothesis
Error df Sig.
df

Value

Pillai's Trace

.032

1.905a

3.000

171.000 .131

Wilks'
Lambda

.968

1.905a

3.000

171.000 .131

Hotelling's
Trace

.033

1.905a

3.000

171.000 .131

Roy's Largest
Root

.033

1.905a

3.000

171.000 .131

Pillai's Trace

.025

1.471a

3.000

171.000 .224

Wilks'
Lambda

.975

1.471a

3.000

171.000 .224

Hotelling's
Trace

.026

1.471a

3.000

171.000 .224

Roy's Largest
Root

.026

1.471a

3.000

171.000 .224

Pillai's Trace

.058

3.495a

3.000

171.000 .017

Wilks'
Lambda

.942

3.495a

3.000

171.000 .017

Hotelling's
Trace

.061

3.495a

3.000

171.000 .017

Roy's Largest
Root

.061

3.495a

3.000

171.000 .017

Pillai's Trace

.009

.493a

3.000

171.000 .688

Wilks'
Lambda

.991

.493a

3.000

171.000 .688

Hotelling's
Trace

.009

.493a

3.000

171.000 .688

Roy's Largest
Root

.009

.493a

3.000

171.000 .688

243

Table 6.72c (Continued)


Multivariate Tests b
Effect
frame_type *
GENDER

Threat_level *
GENDER

frame_type *
Threat_level *
GENDER

Hypothesis
Error df Sig.
df

Value

Pillai's Trace

.017

.966a

3.000

171.000 .410

Wilks'
Lambda

.983

.966a

3.000

171.000 .410

Hotelling's
Trace

.017

.966a

3.000

171.000 .410

Roy's Largest
Root

.017

.966a

3.000

171.000 .410

Pillai's Trace

.001

.041a

3.000

171.000 .989

Wilks'
Lambda

.999

.041a

3.000

171.000 .989

Hotelling's
Trace

.001

.041a

3.000

171.000 .989

Roy's Largest
Root

.001

.041a

3.000

171.000 .989

Pillai's Trace

.047

2.834a

3.000

171.000 .040

Wilks'
Lambda

.953

2.834a

3.000

171.000 .040

Hotelling's
Trace

.050

2.834a

3.000

171.000 .040

Roy's Largest
Root

.050

2.834a

3.000

171.000 .040

a. Exact statistic
b. Design: Intercept + frame_type + Threat_level + GENDER + frame_type *
Threat_level + frame_type * GENDER + Threat_level * GENDER + frame_type *
Threat_level * GENDER

244

Table 6.73d Experiment 3: Hypothesis 8c: tests of between-subjects


effects (mobile phone stimuli)

Source

Corrected
Model

Intercept

Dependent
Variable

Mean
Square

Sig.

Partial
Noncent. Observed
Eta
Parameter Powerb
Squared

Perceived
severity

16.307a

2.330

2.867

.007

.104

20.066

.917

Perceived
vunerability

4.223c

.603

.454

.866

.018

3.178

.196

Fear

19.299d

2.757

1.588

.142

.060

11.119

.649

Perceived
severity

5760.417 1 5760.417 7088.373 .000

.976

7088.373

1.000

Perceived
vunerability

4044.202 1 4044.202 3043.185 .000

.946

3043.185

1.000

Fear

2893.106 1 2893.106 1666.937 .000

.906

1666.937

1.000

Threat_level Perceived
severity

3.036

3.036

3.736

.055

.021

3.736

.485

Perceived
vunerability

.893

.893

.672

.414

.004

.672

.129

Fear

.006

.006

.003

.954

.000

.003

.050

1.223

1.223

1.505

.222

.009

1.505

.230

Perceived
vunerability

.057

.057

.043

.836

.000

.043

.055

Fear

3.577

3.577

2.061

.153

.012

2.061

.298

Perceived
severity

4.382

4.382

5.392

.021

.030

5.392

.637

Perceived
vunerability

.005

.005

.004

.950

.000

.004

.050

12.505

12.505

7.205

.008

.040

7.205

.761

.559

.559

.688

.408

.004

.688

.131

1.563

1.563

1.176

.280

.007

1.176

.190

.023

.023

.013

.909

.000

.013

.051

.063

.063

.077

.782

.000

.077

.059

Perceived
vunerability

.003

.003

.002

.962

.000

.002

.050

Fear

.145

.145

.084

.773

.000

.084

.060

Frame_type Perceived
severity

GENDER

Type III
Sum of df
Squares

Fear
Threat_level Perceived
*
severity
Frame_type Perceived
vunerability
Fear
Threat_level Perceived
* GENDER severity

245

Table 6.73d (Continued)

Source

Dependent
Variable

Frame_type Perceived
* GENDER severity

.075

.093

.761

.001

.093

.061

Perceived
vunerability

.776

.776

.584

.446

.003

.584

.118

Fear

2.937

2.937

1.692 .195

.010

1.692

.253

6.680

6.680

8.220 .005

.045

8.220

.814

.999

.999

.752

.387

.004

.752

.139

.406

.406

.234

.629

.001

.234

.077

Fear

Total

Corrected
Total

Partial
Noncent. Observed
Sig. Eta
Parameter Powerb
Squared

.075

Threat_level Perceived
* Frame_type severity
* GENDER
Perceived
vunerability

Error

Type III
Mean
Sum of df
Square
Squares

Perceived
severity

140.590 173

Perceived
vunerability

229.906 173 1.329

Fear

300.256 173 1.736

Perceived
severity

5920.000 181

Perceived
vunerability

4301.333 181

Fear

3207.560 181

Perceived
severity

156.896 180

Perceived
vunerability
Fear

.813

234.129 180
319.554 180

a. R Squared = .104 (Adjusted R Squared = .068)


b. Computed using alpha = .05
c. R Squared = .018 (Adjusted R Squared = -.022)
d. R Squared = .060 (Adjusted R Squared = .022)

246

Plots were produced to check the interaction effects among the


three factors. The following figures show the various interactions. Figure 6.3
show that women perceive the environmental threat to be more severe to their
health when compared to men. Fear arousal is also greater in women when
compared to men (Figure 6.4). It can be seen from Figure 6.5 that women
perceived high severity in both the low and high threat conditions when gain
framing is used. However men perceived higher levels of severity only under
high threat conditions when gain frames are used. Figure 6.6 shows that under
loss frame condition, higher threat levels evoke higher levels of perceived
severity only in the case of women. Men do not perceive greater threat
severity under loss conditions even when high threat levels are used.

Figure 6.3 Estimated marginal means for perceived severity

247

Figure 6.4 Estimated marginal means for fear

Figure 6.5

Estimated marginal means for perceived severity for gain


frames

248

Figure 6.6

Estimated marginal means for perceived severity for loss


frames

Table 6.74 shows the effect of the manipulations.


Table 6.74

Hypothesis

Experiment 3: Summary of the effect of manipulations on


PMT variables with the mobile stimulus
Factor

Perceived
severity

Perceived
vulnerability

Fear

H5

Threat level

NA

H6

Goal frame

NA

H7

Threat level * Goal Frame

NA

H8c

Gender * Goal Frame

Effect of Interactions on Purchase Intentions


Purchase Intention
H8a

Environmental concern * Goal Frame

H8b

Message Involvement * Goal Frame

249

6.5.8.3

Hypotheses tests of the relationship among PMT variables,


involvement, attitudes and intentions using the structural/ inner
model
As the measurement model was reliable and valid, the structural

model was tested with SmartPLS version 2.0.M3 (Ringle et al 2005). Initially
collinearity among the exogenous constructs was examined and there were no
multicollinearity issues. The main evaluation criterion for the structural model
is the value of the coefficient of determination (R2) as it represents the
explained variance of all the endogenous variables (Hair et al 2011). The level
and significance of the path coefficients (Hair et al 2011) are other important
criteria to judge the model.
The structural model was tested with 5000 sub-samples generated
using bootstrapping to evaluate the significance of the path co-efficients (181
observations per subsample, 5,000 subsamples and no sign changes). The
results of the structural model are shown in Figure 6.7. The significance of the
hypotheses were evaluated based on two-tailed tests (p < 0.05 (t=1.971), p <
0.01 (t= 2.598) and p < 0.001 (t= 3.334)).
The R2 values (shown in brackets) and path coefficients can be seen in Figure
6.7. R2 values greater than 0.33 are substantial and values between 0.19 and
0.33 are moderate (Chin 1998; Henseler et al 2009). Hair et al (2011) suggest
that 0.20 can be considered high for consumer behaviour studies.
Based on Chins criteria (Chin 1998) it can be observed that:
The coefficient of determination, R2 is 0.22 for the fear
endogenous latent variable. This means that the three latent
variables (perceived severity, perceived vulnerability, and

250

environmental knowledge) moderately explain 22.0% of the


variance in fear.
environmental
knowledge
-0.22***

-0.11
-0.04

environmental
concern

fear (0.22)
0.29***
0.35***

perceived
severity

0.07
0.04
message
involvement (0.36)

0.16*
perceived
vulnerability

0.03

0.07
0.56***
0.22**
0.11

0.24*
*

response
efficacy

0.02
0.71***

(0.30)

0.25*
*

attitude towards
ad (0.32)

0.07

0.40***

attitude towards
brand (0.50)
self efficacy
(0.07)

-0.01
0.18*

0.62***
-0.01
purchase
intention (0.46)

Figure 6.7 Hypothesis testing using PLS-SEM


The coefficient of determination, R2 is 0.36 for the message
involvement endogenous latent variable. This means that the six
latent variables. (perceived severity, perceived vulnerability,
response efficacy, self efficacy, fear and environmental

251

concern) substantially explain 36.0% of the variance in message


involvement.
The coefficient of determination, R2 is 0.32 for the attitude
towards ad endogenous latent variable. This means that the four
latent variables (response efficacy, self efficacy, message
involvement and environmental concern moderately explain
32.0% of the variance in attitude towards ad.
The coefficient of determination, R2 is 0.50 for the attitude
towards brand endogenous latent variable. This means that the
the latent variable (attitude towards ad) substantially explains
50.0% of the variance in attitude towards brand.
The coefficient of determination, R2 is 0.46 for purchase
intention. This means that the latent variable - attitude towards
brand substantially explains 46.0% of the variance in purchase
intention.
Based on the inner model loadings and path co-efficients from
Figure 6.7, it can be summarised that:
The hypothesized path relationship between perceived severity,
perceived vulnerability and fear is statistically significant.
perceived severity has a comparatively stronger effect (0.29) on
fear.
perceived severity (0.40) and perceived vulnerability (0.24)
significantly predict response efficacy
The hypothesized relationship between perceived severity and
self efficacy was significant (0.25) whereas perceived
vulnerability was not significantly related to self efficacy (0.07)

252

The hypothesized path relationship between perceived severity


(0.04), perceived vulnerability (0.07), self efficacy (0.11) and
message involvement was not statistically significant. However
fear (0.35) and response efficacy (0.22) had a signficiant
relationship with message involvement.
response efficacy (0.07) and self efficacy (0.02) were not related
to attitude towards ad. While response efficacy (-0.01) was not
related to purchase intention, the hypothesized relationship
between self efficacy (0.18) and purchase intention was
significant.
The hypothesized path relationship between perceived severity
(-0.11), perceived vulnerability (-0.04), message involvement
(0.07) and environmental knowledge was not statistically
significant. However fear (-0.22) had a signficiant relationship
with environmental knowledge
environmental concern was not related to any of the
hypothesized relationships. The path coefficients were not
significant

with

the

hypothesized

variables

message

involvement (0.07), attitude towards ad (0.03) and purchase


intention.
message involvement (0.56) is a significant predictor of attitude
towards ad
attitude towards ad is a strong predictor (0.71) of attitude
towards brand
Similarly attitude towards brand is a significant predictor (0.62)
of purchase intention.

253

Table 6.75 summarises the result of the hypotheses tests and their associated
results.
Table 6.75 Experiment 3: Results of hypothesis testing using PLS-SEM
Hypothesis
H9a
H9a

Path
Perceived severity

fear

Perceived vulnerability

fear

response

(standard)

tvalue

0.29 ***

4.03

Supported

2.02

Supported

5.92

Supported

0.16

***

H9b

Perceived severity
efficacy

H9b

Perceived vulnerability
response efficacy

0.24 **

3.03

Supported

H9c

Perceived severity
efficacy

0.25 *

2.50

Supported

H9c

Perceived vulnerability
efficacy

0.07

0.58

Not
supported

H10a

Perceived severity
involvement

0.04

0.50

Not
supported

H10b

Perceived vulnerability
message involvement

0.07

0.89

Not
supported

H10c

Fear

0.35***

4.82

Supported

H10d

Response efficacy
involvement

0.22**

2.63

Supported

H10e

Self efficacyinvolvement

0.11

1.57

Not
Supported

H11a

Response efficacy
towards ad

0.07

0.92

Not
Supported

H11b

Self efficacytowards ad

0.02

0.35

Not
Supported

H11c

Response efficacy
intention

-0.01

0.22

Not
Supported

self
self

message

message involvement
message

message
attitude

attitude
purchase

0.40

Result

254

Table 6.75 (Continued)

(standard)

tvalue

purchase

0.18*

2.52

Supported

H12a

Environmental knowledge
perceived severity

-0.11

1.45

Not
Supported

H12b

Environmental knowledge
perceived vulnerability

-0.04

0.62

Not
Supported

H12c

Environmental knowledge

-0.22***

3.47

Supported

Hypothesis
H11d

Path
Self efficacy-

Result

intention

fear
H12d

Environmental knowledge
message involvement

-0.08

1.349

Not
Supported

H13a

Environmental concern
message involvement

0.07

1.08

Not
Supported

H13b

Environmental concern
attitude towards ad

0.03

0.46

Not
Supported

H13c

Environmental concern

-0.01

0.316

Not

purchase intention
H14

Message involvement
attitude towards ad

H15

Attitude towards ad
towards brand

H16

Attitude towards brand

Supported

attitude

0.56 ***

9.50

Supported

0.71 ***

17.7

Supported

0.62 ***

10.3

Supported

purchase intention
Note: n=181; Estimates represent 5000 bootstrapping testing
*p<0.05 ; **p < 0:01; * **p <0:001

255

It is also necessary to assess the predictive relevance of the inner


model and therefore the models predictive relevance was analyzed using
Stone-Geisser test criterion Q2 (Chin. 2010; Hair et al 2011). This was
determined using the blindfolding procedure in SmartPLS. The omission
distance was chosen as 7, since values between 5 and 10 are advantageous
(Hair et al 2012). Cross-validated measure Q2 was checked and the results are
shown in Table 6.76. The Q2 values for all the endogenous constructs were
greater than zero as required. Table 6.76 also lists the R2 values of the
endogeneous contructs.
Table 6.76 Experiment 3: Models predictive relevance
Endogenous Construct

R2

Q2

Attitude towards ad

0.32 0.23

Attitude towards brand

0.50 0.43

Purchase intention

0.46 0.40

Message involvement

0.36 0.19

Fear

0.22 0.15

Next, the effect size was calculated to measure the impact of a


predictor on a specific endogenous construct. Effect size represents Cohens d.
The values of 0.02 (small), 0.15 (medium) and 0.35 (large) indicate that the
construct has a small, medium or large effect size on the criterion (dependent)
construct respectively (Cohen 1988). The effects are shown in Table 6.77.
Although fear has a small effect on message involvement the effect size 0.14
is very close to the medium threshold. Most of the other effect sizes are small.
However message involvement has a medium effect on attitude towards ad
and attitude towards brand has a medium effect on purchase intention.
Attitude towards ad has a strong effect on attitude towards brand.

256

Table 6.77 Experiment 3: Effect sizes


Exogenous

Endogeneous Variable

Variable

Effect
Size f2

Effect size
Interpretation

Purchase intention

Attitude Towards brand

0.32

Medium

Purchase intention

Self efficacy

0.05

Small

Attitude towards brand

Attitude towards ad

0.43

Large

Attitude towards ad

Message involvement

0.26

Medium

Attitude towards ad

Response efficacy

0.01

Small

Attitude towards ad

Environmental concern

0.06

Small

Message involvement

Fear

0.14

Small

Message involvement

Perceived vulnerability

0.01

Small

Message involvement

Response efficacy

0.04

Small

Message involvement

Self efficacy

0.01

Small

Message involvement

Environmental concern

0.01

Small

Fear

Perceived severity

0.09

Small

Fear

Perceived vulnerability

0.03

Small

Fear

Environmental knowledge

0.06

Small

Perceived severity

Environmental knowledge

0.01

Small

6.5.9

Conclusions from Experiment 3


This experiment examined the effects of varying threat levels and

goal frames on PMT variables and the subsequent effects of the PMT
variables on message involvement, attitudes and purchase intention. The issue
of e-waste (based on mobile phone stimuli) was used to observe if the
hypothesized effects were supported. The levels of perceived severity and
vulnerability remained high in this experiment too. It can be inferred from
Table 6.67 and 6.68 that perceived severity, vulnerability and fear remain

257

high for watch stimuli. Similar to Experiment 2 there was no main effect of
threat level or the goal frames. Similarly the hypothesized interaction effects
were only partially significant. Similar to Experiment 2 loss frames did not
increase threat perception as suggested by other researchers (Cox & Cox
2001; Meyers-Levy & Maheswaran 2004; van t Riet et al 2008;OKeefe &
Jensen 2009; Janssens et al 2010; Updegraff 2013).
There was no relationship between involvement and framing as
hypothesized. Both environmental concern and message involvement did not
interact with frames to produce an effect on purchase intention. This is in
contrast to the findings by other researchers who imply an effect between
involvement and framing (Maheswaran & Meyers-Levy 1990; Rothman et al
2006; Kim 2013). However gender played a significant role in predicting the
effect of the factors on the PMT variables. Women were more fearful and
perceived higher severity when facing an environmental threat when
compared to men. This confirms earlier findings by Garbarino & Strahilevitz
(2004) who find that women are more risk averse when compared to men.
The result also highlights the gender gap known to exist in environmental
threat perceptions (Flynn et al 1994; Bord & Connor 1997; McCright &
Dunlap 2011; Franzen & Vogl 2013). The interaction between frame and
gender also showed that women generate more negative thoughts when
presented with a negative goal frame when compared to men (Putrevu 2010).
Similar to the previous experiment, perceived severity and
vulnerability significantly influenced fear arousal as proposed by PMT
(Rogers & Prentice-Dunn 1997 ; Floyd et al 2000) and other studies that
apply this theory (Milne et al 2000; de Hoog et al 2008). Coping appraisal
was also initiated (Maddux & Rogers 1983; Boer & Seydel 1996; Milne et al
2000). Response efficacy was significantly influenced by perceived severity
and vulnerability. However self efficacy was only moderately influenced by

258

perceived severity and not by perceived vulnerability. In this experiment too,


all the PMT variables except self-efficacy predicted message involvement.
This finding confirms that higher levels of health risk increase involvement
with the message (Bloch & Richins 1983; Richins & Bloch 1986; Keller &
Block 1996; de Hoog 2005). Fear can therefore be used to increase message
involvement in enviromental communication (Meijnders et al 2001).
Apart from this, the finding also shows that while using goal frames
and threat levels, emphasis on the response efficacy would significantly
increase consumer involvement. The results are also in contrast to the finding
by Punam & Keller (1995) who find that low efficacy promotes more effortful
processing. Recent research emphasizes the importance of response efficacy
and treats it as a key component to message acceptance (Lewis et al 2010).
However, contrary to previous research, the efficacy variables were not
related to attitudes. Self efficacy significantly predicted purchase intentions
confirming the findings of earlier research (Maibach & Murphy 1995;
Luszczynska 2004; Gaston & Prapavessis 2012; Kreausukon et al 2012).
While objective environmental knowledge did not decrease the levels of
perceived severity or vulnerability, it had a negative effect on fear (Averbeck
et al 2011). Knowledge did not affect the levels of message involvement
similar to Experiment 2. Interestingly, environmental concern did not have an
effect on any of the hypothesized variables. This is in contrast to the findings
of Experiment 2 and contrary to the ELM (Petty & Cacioppo 1986). However ,
this finding was related to Experiment1 where environmental concern did not
predict the hypothesized variables. From Table 6.77, it can also be inferred
that message involvement has a stronger effect on attitude towards the ad.
Hence, this experiment based on stimuli related to e-waste showed
similar results as Experiment 2. The major difference was the role of
environmental concern. While Experiment 2 showed an effect of

259

environmental concern, this experiment showed that this variable was not
effective in influencing attitudes or intentions.

Apart from the findings

related to environmental concern, there were no major differences between


the issue of e-waste and plastic waste.There were only two other differences.
In case of Experiment 2, perceived severity affected message involvement
and vulnerability affected self-efficacy. Hence, both the experiments highlight
the role of message involvement in promoting attitudes and intentions towards
green advertising.
6.5.10

Gender and Environmental Concern


Some green marketing studies claim that gender has a significant

effect on environmental concern (Shrum et al 1995; Jain & Kaur 2006;


Mostafa 2007). This was not investigated as part of hypothesis testing as it
was not part of the research objective, Surprisingly, a post hoc analysis
revealed that gender did not have any significant effect on environmental
concern in all the three experiments. This confirms the recent finding by other
researchers who do not find a link between environmental concern and gender
(e.g. DSouza et al (2007)).

Vous aimerez peut-être aussi