A cross-national investigation into the relation between advertising explicit
memory, advertising likeability, and consumer behavior intentions
Feng Hu
University of Groningen, Groningen, the Netherlands
University of Chinese Academy of Sciences, Beijing, China
Abstracts: Advertising likeability has been heavily focused on by academia and marketers.
And its effects on advertising memory and consumer behavior intentions are also
examined in former studies. However, these effects may be influenced by cultural factors,
and the robustness of these effects across countries is not well elaborated on. In addition,
there lacks research on the effect of advertising memory on advertising likeability so far.
This study aims at these gaps. In this paper, the author investigates cause-and-effect
relationships between advertising explicit memory, advertising likeability, and two kinds
of consumer behavior intention: purchasing and Word-of-Mouth. We use the cross
national sample data collected by MetrixLab in six countries, and conduct analyses using
OLS, HLM, and mediation tests. The results show that advertising likeability carries the
influence of advertising memory on both kinds of consumer behavior intention. And
advertising likeability can also influence advertising memory through the linear and
nonlinear way. These findings are quite consistent among six countries. We also find the
dual relationships between advertising memory and advertising likeability. Limitations
and future research avenues are also provided in the last section.
Keywords: advertising likeability, advertising explicit memory, purchasing intention,
Word-of-Mouth
1. Introduction
2 / 59
are positive effects of advertising likeability on advertising effectiveness (D. Aaker &
Stayman, 1990), advertising attitude, attitude shift, recommendation (Smit et al., 2006),
persuasion (Alexander L. Biel & Bridgwater, 1990), intention to purchase (Fam, 2008),
and brand attitude (Walker & Dubitsky, 1994). With the rapid development of
communication technology, consumers will receive an abandon of information in the
daily. Because of the limitation of human being’s mental capacity to process the
information, consumers can not carefully process every piece of information. So they
increasingly turn to depend on some rule of thumbs to decide which information should
be paid much attention. In this situation, advertising which is likeable and relevant will
predominate over other stimulus, and have the great chance to be noticed by consumers.
So, in the current media environment, it should be more crucial to create likeable and
relevant advertising (Smit et al., 2006).
Not only does advertising likeability affect consumer behaviors, but it can influence
cognitive processing of advertising messages. Studies in this field show that a well
likeable advertising can affect information processing by creating positive arousal and
activation, improving advertising recall, and further producing positive judgment of
message (D.A. Aaker, Batra, & Myers, 1992; Burke & Edell, 1986; Hwiman & Xinshu,
2003; Smit et al., 2006; Tao, Wells, Xinshu, & Seounmi, 2001).
the more the consumer retain the advertising, the more the advertising is likeable.
However, studies in advertising likeability seldom focus on this reversed logical sequence,
which is from advertising memory to advertising likeability.
This paper aims at both research gaps. Firstly, the relationships between advertising
memory, advertising likeability, and consumer behavior intentions will be investigated in a
more diversified cultural setting. Comparing to the studies with the same research field,
findings in this article will be more generalized. Secondly, the effect of advertising
memory on advertising likeability will be tested using regression and multiple mediating
tests under the guide of knowledge of meta-cognition. Indeed, advertising likeability
carry substantial influence of advertising memory on consumer behavior intentions.
Besides these two major points, the nonlinear effect of advertising likeability on
advertising memory will be explored. We find that there is U-shape relationship between
“easy-to-understand”, one of advertising likeability measurements, and advertising
memory. This finding extends our knowledge of the linear effect of advertising
likeability on advertising memory.
We organize the rest of this article as follows: we present our conceptual model, and
generate the hypotheses concerning the relationship between advertising memory,
advertising likeability, and consumer behavior intentions. Then, we discuss the sample,
data collection procedure, measurement of the included variables, and econometric
model and describe the empirical results. In addition, the robust checks will be provided.
Finally, we present the theoretical and management implications in general discussion
section and provide the limitations and future research avenues in the last section.
2. Conceptual Model
4 / 59
and message-recall. Thus, advertising memory here is a kind of measure for explicit
memory (Fennis & Stroebe, 2010). In this article, “advertising memory” and “advertising
explicit memory” are exchangeable.
Figure 1 Conceptual Model
Consumer behavior
·
Purchasing
intention
·
Word-of-Mouth
Ad Likeability:
affective part
·
Ad like
·
Ad interesting
·
Ad credible
·
Ad irritating
·
Ad fitness
·
Ad remarkable
·
Ad relevance
Ad Likeability:
cognitive part
·
Easy-to-understand
Linear U-shape Linear U-shapeAd explicit memory
·
Brand-recall
·
Message-recall
Not significant3. Theoretical Reasoning & Hypotheses
In this section, theoretical reasoning behind the conceptual model well be presented,
and three main hypotheses will be generated, which include the effects of advertising
likeability on purchase intention and Word-of-Mouth, the mediating effect of advertising
likeability who carry the influence of advertising memory on purchase intention and
Word-of-Mouth, and the effect of advertising likeability on advertising memory.
Effects of advertising likeability on purchase intention and Word-of-Mouth
If the advertising is likeable, consumers may be willing to take note of it, and may be
willing to watch it again (A.L. Biel, 1990; Smit et al., 2006). And if consumers prefer
more to a commercial, they are more likely to continue and process it thoroughly. In this
case, consumers are more willing to tell someone else this “good” commercial, and they
are more likely to be persuaded by this commercial and buy the product. Other empirical
studies also find the positive effects of advertising likeability on purchase intention (e.g.
(Fam, 2008)) and recommendation (e.g. (Smit et al., 2006)). In addition, adverting itself
can be looked as a brand attribute (A.L. Biel, 1990). This is especially true when function
characters of different brands are perceived as very similar in the same categories. So,
liking the brand will be correlated to some positively actions towards brands, such as
purchasing the brands or Word-of-Mouth (Smit et al., 2006). Finally, a likeable advertising
can induce positive emotions, which lead to positive attitudes towards the advertising,
and furthermore, positive attitudes towards the brand. And those positive attitudes will
results in a higher purchase intention and Word-of-Mouth (Smit et al., 2006). Thus, we
expect the following:
H1: there is a positive effect of advertising likeability on purchase intention and
Word-of-Mouth.
The mediation effect of advertising likeability between advertising memory and purchase
intention & Word-of-Mouth
6 / 59
Consumers cannot choose a brand, unless they are aware of the existence of the brand
(Lavidge & Steiner, 1961). However, the direct effect of advertising memory on
consumer behaviors is quite trivial. Here are the reasons. First of all, Advertising
memory used in this article is a kind of explicit memory (Fennis & Stroebe, 2010).
However, some empirical studies find that in some situations, implicit memory is a more
reliable predictor for attitudes towards the brand, not explicit memory (Shapiro &
Krishnan, 2001; Yoo, 2008). So, in this paper, only explicit memory may not well predict
the attitude towards the brand, and further, the consumer’s actions. Furthermore, other
studies reveal that there are two kinds of original information stored in consumer’s brain:
the information about the attribute of the products and the consumer’s evaluation based
on their elaboration (Carlston, 1980; Kardes, 1986). Obviously, measures of advertising
memory used in this paper belong to the information about the attribute of the products.
However, it is the consumer’s evaluation based on their elaboration that is retrieved for
use in making subsequent global memory-based judgments. And it is the global
judgments that predict purchase intentions. So, measures of advertising memory are poor
measures of advertising effectiveness (Fennis & Stroebe, 2010). Thus, we express the
hypothesis as follows and expect that this hypothesis will be invalid.
H2a: there is direct influence of advertising memory on purchasing intention
and Word-of-Mouth.
meta-cognation. Ease of retrieval can be defined as the apparent ease with which
product and brand related information can be retrieved from memory (Fennis & Stroebe,
2010). Consumer may infer the validity of retrieved information based on the ease of
retrieval: information which can be more easily retrieved should be more correct. Other
studies also reach the similar conclusion that consumer will be convinced by the recall
information if the ease of recall increase or the difficulty of recall decrease (Haddock et
al., 1999; Haddock et al., 1996; Nelson & Narens, 1990; Tormala et al., 2002). Fennis &
Stroebe (2010) further concludes that information which can be easily retrieved may
positively affect product attitude. In this paper, the retrieved information is measured by
advertising memory, which is aggregated by two variables: message recall and brand recall.
Message recall is an overall evaluation of the advertising memory, while brand recall just
covers the brand part of advertising memory. So, the positive effect of brand recall on
advertising likeability may be weaker than message recall. We will check it later in “robust
checks” section. In conclusion, based on meta-cognition knowledge, we can safely
conjecture that the advertising will be more likeable if information about this advertising
can be more easily recalled. Thus, we expect the following:
H2b: there is a positive influence of adverting memory on advertising
likeability.
Combing the H1, H2a and H2b, we expect the following:
H2: advertising likeability mediates the influence of advertising memory on
purchasing intention and Word-of-Mouth.
Effect of advertising likeability on advertising memory
8 / 59
which can be retrieved later (Lingle & Ostrom, 1981; Zajonc, 1980; Zinkhan, Locander,
& Leigh, 1986).
A likeable advertising can create positive arousal and activation, which will affect
information processing, (Kroeber-Riel, 1979). A likeable advertising can improve the
recall of the advertising: the more the advertising is likeable, the more likely the
consumer will remember it (D.A. Aaker et al., 1992). Other studies use memory encoding
to explain the influence of liking on memory. Cues can play roles in what is encoded, in
how the information is stored, and later, retrieved (Tulving & Thomson, 1973). Liking
towards the advertising can be used as cues by consumers to encode, store, and retrieve
the advertising information (Tao et al., 2001).
And at the other point of the continuum of liking, if the advertising is less likeable,
consumers may avoid processing the advertising information, or try to ignore this
advertising (van Raaij, 1989). Thus, we expect the following:
H3a: there is positive influence of advertising likeability on advertising
memory.
If information contained in the advertising will be retained by consumers, consumers
have to pay attention to the information. Attention is a necessary condition for memory.
Factors that affect attention are numerous, and have been heavily studied. In general,
attention literatures propose that something which is more distinctive or unexpected, will
receive greater attention (Houston, Childers, & Heckler, 1987; Wallace, 1965). And a
former study also shows that when an advertising message contained information
incongruent with a person's general knowledge structure about a domain (i.e., schema),
processing of the entire message was enhanced (Hunt, Kernan, & Bonfield, 1992).
Consumers’ high attention can be achieved by violating their expectations of the
information (Votolato, Montgomery, & Unnava, 2007).
H3b: there is a “U-shape” relationship between advertising likeability and
advertising memory.
4. Research methodology
Data collection
Sample data used for testing the hypotheses are collected by MetrxLab, which is a
Europe's online research agency. These sample data come from 6 countries: Australia,
Russia, China, Poland, Mexico, and Japan. There are 28 survey projects in these data.
Specifically, China and Russia contain 4 projects respectively, and the rests countries
contain 5 projects respectively. In each project, averagely 290 respondents were recruited,
and projected with 8 pieces of commercials. Then they were asked to evaluate these
commercials at the same time. Both the commercial projection and evaluation were
conducted on the computer. Each project contains different respondents. On average,
there are 2934 valid observations in each country.
Survey measurement
Variables used in this article are measured in Likert scale. The details of the survey
measurement can be found in table 1.
10 / 59
The construct of advertising memory is measured by brand-recall and message-recall.
Consumer behaviors in this conceptual model consist of 2 kinds: purchasing intention
and Word-of-Mouth.
Table 1 Survey measurement
Construct
Items
Advertising
likeability:
cognitive part
Easy-to-understand: What do you think of this ad? 1- Difficult to
understand ~ 5- Easy to understand.
Advertising
likeability:
affective part
Like:
What do you think of this ad? 1- Do not like the ad ~ 5- Like the ad.
Interesting: this ad is interesting. 1- Totally disagree ~ 5- Totally agree.
Credible: What do you think of this ad? 1- Not credible ~ 5- Credible.
Irritating: What do you think of this ad? 1- Irritating ~ 5- Not irritating
Fitness: What do you think of this ad? 1- Do not fit its brand / product ~
5- Fit its brand / product
Remarkable: What do you think of this ad? 1- Unremarkable (same as
others ads) ~ 5- Remarkable (different from other ads)
Relevance: What do you think of this ad? 1- Not relevant to my needs ~ 5-
Relevant to my needs.
Consumer
behavior
Purchasing intention: I want to buy this product once I see this
advertising. 1- Totally disagree ~ 5- Totally agree.
Word-of-mouth: How likely are you to forward this video to your
friends or family (e.g. per Email)? 1 - Very unlikely ~11 - Very likely
Advertising
memory
Brand-recall: Please indicate below for which brands you have seen a
commercial when watching the TV-commercials 1-no; 2-maybe; 3-yes.
Message-recall: Can you remember what this advertising wanted to
make clear to you with this advertisement? 1-no; 2-partially; 3-yes.
Validity and reliability of measures
The coefficient alphas of advertising likeability: affective part and advertising memory
are 0.88 and 0.59 respectively. Although for advertising memory, the coefficient alpha is
not high enough, 0.59 is also acceptable for marketing research (e.g. (Verhoef & Leeflang,
2009))
1. We further assess the reliability and validity of the scales using exploratory and
confirmatory factor analysis for advertising likeability: affective part and advertising
memory separately. In exploratory factor analysis, the loadings per item per construct are
all greater than 0.69. And only one component can be aggregated for advertising
likeability: affective part and advertising memory. In confirmatory factor analysis, all
standardized factor loadings are significant (p<0.001) and greater than 0.6
2.
Data description
On average, there are 23% of brands belonging to service industry and the rest are all
product brands. These brands are mainly distributed in sector of electronics, FMCG
nonfood, FMCG food and Automobile. For respondents, over 70% of respondents are
18 - 45 years old. Over 50% of respondents attend some college or got some degree.
Nearly 75% of respondents have full or part time jobs. And over half of respondents are
married or living together. Details of sample descriptions per country can be found in
Table 2
Table 2 sample descriptions per country
Australia Russia China Poland Mexico Japan
Advertising related
Survey projects 5 4 4 5 5 5
Brands 40 32 32 40 40 40
Product or service (percentage)
Service 34 22 25 18 24 16 Product 66 78 75 82 76 84 Valid observations 3089 2402 2610 3768 3032 2713 Sector (percentage) electronics 25 20 29 35 23 36 FMCG Nonfood 10 20 3 18 5 9 FMCG food 13 27 31 13 33 15 automobile 5 10 12 10 16 11 telecom 10 7 9 7 5 2 finance 7 10 6 7 11 10 retailer 2 3 3 2 2 0 Leisure-related 5 0 0 0 5 9 utilities 2 0 0 0 0 5
clothing & jewelry 0 0 0 3 0 4
other (reference) 20 3 6 5 0 0 Respondent related Gender (percentage) Female 56 54 35 50 42 46 Male 44 46 65 50 58 54 Age (percentage)
2 Because the specific structure of the sample data, the variance-covariance matrix of the estimates for confirmatory
12 / 59 18 to 25 years 9 20 8 15 12 4 26 to 30 years 9 22 21 15 13 8 31 to 35 years 9 18 28 11 14 13 36 to 40 years 10 15 17 15 15 15 41 to 45 years 12 10 13 11 15 16 46 to 50 years 13 6 7 9 11 16 51 to 55 years 13 5 3 10 20 13 56 to 60 years 13 1 2 8 0 8 61 to 65 years 13 0 0 5 0 6 Education (percentage)
Attended or completed some vocational /
professional training 24 21 4 3 44 14
Attended some college or got some degree 40 69 90 56 38 60 Less than high school or high school graduate
(reference) 35 10 6 41 17 26
Occupation (percentage)
working in full or part time 56 80 95 72 77 74
study / education 5 6 3 8 7 2
retired, or disabled, or housewife / man 33 12 2 14 11 19
unemployed (reference) 6 3 0 5 6 5
Household situation (percentage)
Living with (grand)parents / relatives 12 12 10 23 18 24 Single, no child or with child(ren) 25 16 8 14 22 19
Student house, commune 1 1 2 3 0 0
Married / living together 54 68 79 53 56 54
other (reference) 7 2 1 6 4 4
Econometric model
According to the conceptual model, we formulate these econometric models as
follows:
0p 1p 2p 3p i ip i pPI
LA
LC
AM
CV
………...… (1)
0w 1w 2w 3w i iw i wWOM
LA
LC
AM
CV
………. (2)
2 2 0m 1m 2m 3m 4m i im i mAM
LA
LA
LC
LC
CV
…… (3)
0la 1la i ila i laLA
AM
CV
……….……… (4)
0lc 1lc i ilc i lcLC
AM
CV
……….………. (5)
likeability: cognitive part, and advertising memory respectively; CV refers to control
variables which consist of product or service, advertising size (second), gender, age,
brand sector, education, occupation, and household situation. Scores of LA and AM are
averaging scores of items which belong to the construct respectively. PI, WOM, LA, LC,
and AM are normalized with mean of 0, and standard deviation of 1.Equation 1 and 2
are established for H1 and H2a; equation 3 is for H3a and H3b; and equation 4 and 5 are
for H2b.
In order to test the consistency of outcomes, the equations are fitted using every
country’s data respectively. The pooled data are also used. Because data in Japan do not
have advertising size variable, there are two versions for pooled data: pooled without
Japan, which means advertising size will be added in the model, and pooled with Japan,
which means advertising size will not be added in the model.
Intra-class Correlation & Multicollinearity
Table 3 ICC
Country
DV
ICC
Country
DV
ICC
14 / 59
Observations in our sample are not independent with each other. And these
correlations are mainly contributed by two points. The first one is that each commercial
is evaluated by about 290 respondents. The same contents of commercials will lead to
the similar evaluations from the respondents. So observations which belong to the same
commercials are correlated with each other. The second one is that every respondent are
asked to evaluate 8 pieces of commercials. Observations which belong to the same
respondents are also correlated with each other. Because of correlations between
observations, simple OLS is not applicable.
In order to deal with the observation correlation, we firstly calculate the intra-class
correlation for the dependent variables in these 5 equations. We set the commercial level
as macro level, and observations level as micro level to calculate the Intra-class
Correlation Coefficients (ICC).
Former literatures recommends that OLS will be applicable if ICC is smaller than 0.1,
and HLM will be more appropriate if ICC is greater than 0.1(e.g. (V. E. Lee, 2000)). We
find that ICCs for PI and WOM are not greater than 0.1 in each country, which means
that correlation resulting from the first source is not severe. Thus only correlations
resulting from the second source should be accounted for. In that case, equation 1 and 2
are estimated by OLS, which is specified to allow for the correlation between
observations which belong to the same respondents, and independent across
observations which do not belong to the same respondents
3. Although in two pooled
data sets, ICCs for PI and WOM are substantial (>0.15), we find that estimation results
which are from OLS are quite similar with their counterparts obtained using Hierarchal
Linear Model. Thus, for simplicity, in equation 1 and 2, we take the outcomes obtained
using OLS as final. We also provide the outcomes obtained using HLM in Appendix for
robust check.
For AM, LA, and LC, we find that their ICCs are greater than 0.1 in at least one
country. Thus, for equation 3, 4, and 5, we build the random intercept model to account
for the heterogeneity between different brands (Wang, Xie, & Fisher, 2011).
We calculate the correlation coefficient matrix for PI, WOM, LA, LC, and AM. There
are 4 correlation coefficients are greater than 0.4. However, 3 of them are correlation
coefficients between dependent and independent variables (0.72, 0.62), or for both
variables never appearing in the same model (0.68) simultaneously. Thus, the majority of
correlation coefficients are less than 0.4, which indicate no severe multicollinearity
problems (Leeflang & Wittink, 2000). We conclude that multicollinearity does not bias
the estimation results.
Table 4 correlation matrix of constructs in model
Variables
PI
WOM
LA
LC
AM
Purchase Intention (PI)
1
Word-of-Mouth (WOM)
0.68
1
Advertising likeability: affective (LA)
0.72
0.62
1
Advertising likeability: cognitive (LC)
0.26
0.22
0.45
1
Advertising memory (AM)
0.12
0.08
0.15
0.07
1
Mediation
Hypothesis 2 entails a mediation analysis for advertising likeability. In this article,
advertising likeability consists of two parts: affective part and cognitive part. Thus, the
parallel multiple mediation analysis is needed. For inference about the indirect effect in
models with these two mediation component, we also calculate the 95% bias-correlated
bootstrap confidence intervals for all indirect effects. The number of bootstrap samples
is 10000. We also consider the heteroscedasticity of the errors in estimation of the
outcome variables, and use the heteroscedasticity-consistent standard error to get the
more efficient coefficient estimation (Andrew F Hayes & Cai, 2007; Long & Ervin, 2000).
Mediation analysis is conducted in PROCESS, which is developed by Andrew F. Hayes
(A.F. Hayes, 2013).
5. Model results
Effects of advertising likeability on purchase intention and WOM
Table 5 effect of advertising likeability on purchasing intention
DV: Purchasing
intention (OLS) (1) AUS (2) RUS (3) CHI (4) POL (5) MEX (6) JAP
(7) pooled without Jap
16 / 59
Ad memory (*100) 0.66 2.37 0.17 0.66 3.87** -3.13 0.27 0.20
Ad likeability: cognitive -0.14*** -0.06*** -0.03 -0.13*** -0.09*** -0.07*** -0.10*** -0.09***
For the estimation results of control variables, please refer to the appendix A
N 3089 2402 2610 3768 3032 2713 14901 17614 adj. R2 0.54 0.60 0.68 0.55 0.55 0.42 0.60 0.59 Max VIF4 5.88 10.05 19.32 6.91 23.15 8.31 6.16 6.43 Mean VIF 2.53 3.54 4.82 2.97 5.09 3.27 2.63 2.72 Min VIF 1.06 1.08 1.04 1.05 1.03 1.04 1.02 1.04 * p < 0.05, ** p < 0.01, *** p < 0.001
Table 6 effect of advertising likeability on WOM
DV: WOM (OLS) (1) AUS (2) RUS (3) CHI (4) POL (5) MEX (6) JAP (7) pooled without Jap (8) pooled with Jap Ad likeability: affective 0.65*** 0.72*** 0.80*** 0.69*** 0.67*** 0.51*** 0.72*** 0.70*** Ad memory (*100) -0.11 3.21 -1.04 4.98*** 1.01 -0.22 0.60 1.49* Ad likeability: cognitive -0.17*** -0.06** -0.06*** -0.14*** -0.05** -0.05* -0.10*** -0.08***
For the estimation results of control variables, please refer to the appendix A
N 3089 2402 2610 3768 3032 2713 14901 17614 adj. R2 0.34 0.48 0.60 0.43 0.45 0.25 0.49 0.48 Max VIF 5.88 10.05 19.32 6.91 23.15 8.31 6.16 6.43 Mean VIF 2.53 3.54 4.82 2.97 5.09 3.27 2.63 2.72 Min VIF 1.06 1.08 1.04 1.05 1.03 1.04 1.02 1.04 * p < 0.05, ** p < 0.01, *** p < 0.001
In the table 5 and 6 (for the overall estimation results, please refer to Appendix A,
here we only present the estimation results of the key variable; for overall estimation
results obtained using HLM, please also refer to Appendix A), we report the estimation
results of equation 1 and 2 regarding the effect of advertising likeability on purchasing
intention and WOM. The results show the importance of advertising likeability: affective
part as determinant of purchasing intention and WOM. The relations between
advertising likeability: affective part and purchasing intention or WOM are quite
significant (p<0.001), magnificent (coefficients > 0.5), and consistent among different
countries.
The direct influence of advertising memory on purchasing intention and WOM is
quite trivial. It only has positive influence on purchasing intention in Mexico, and on
4 For simplicity, we do not provide all the VIFs, but just the MAX, mean, and MIN VIFs. In fact, this MAX VIFs in
WOM in Poland and Pooled data. Comparing to advertising likeability: affective part, the
magnitude of effects of advertising memory on both DVs are quite smaller.
However, for advertising likeability: cognitive part, the findings in table 5 and 6 are
quite contradicted with our hypothesis 1. There are consistent and significant negative
relationships between advertising likeability: cognitive part and two DVs (marginal
significant in purchasing intention model in China). We will provide some potential
explanations for this negative effect of advertising likeability: cognitive part in following
general discussion section.
In summary, our estimation results demonstrate the significant and consistent positive
effect of advertising likeability: affective part on purchasing intention and WOM.
However, its cognitive part has significant and consistent negative influence on both DVs.
Thus, H1 is partly supported. We also find the direct impact of advertising memory on
both DVs is not consistent, and the effect of advertising memory is quite trivial. Thus,
H2a is strongly not supported, just as we expect.
The mediation effect of advertising likeability
Table 7 effect of advertising memory on Ad likeability: affective part
DV: likeability-affective
(HLM) (1) AUS (2) RUS (3) CHI (4) POL (5) MEX (6) JAP
(7) pooled without Jap
(8) pooled with Jap
Ad memory 0.14*** 0.05* 0.18*** 0.13*** 0.11*** 0.07*** 0.13*** 0.12***
For the estimation results of control variables, please refer to the appendix B
N 3089 2402 2610 3768 3032 2713 14901 17614
Log restricted-likelihood -4189.74 -3309.77 -3518.29 -5231.16 -4012.00 -3767.55 -19693.30 -23332.91
* p < 0.05, ** p < 0.01, *** p < 0.001
Table 8 effect of advertising memory on Ad likeability: cognitive part
DV: likeability-cognitive
(HLM) (1) AUS (2) RUS (3) CHI (4) POL (5) MEX (6) JAP
(7) pooled without Jap
(8) pooled with Jap
Ad memory 0.14*** 0.03 0.10*** 0.07*** 0.05** 0.04 0.08*** 0.08***
For the estimation results of control variables, please refer to the appendix B
N 3089 2402 2610 3768 3032 2713 14901 17614
Log restricted-likelihood -4155.20 -3187.59 -3634.86 -5176.03 -4218.64 -3746.16 -20226.75 -23387.26
* p < 0.05, ** p < 0.01, *** p < 0.001
18 / 59
display the estimation results for the effect of advertising memory on advertising
likeability. In table 7, we find that the effect of advertising memory on advertising
likeability: affective part is quite significantly positive and consistent between different
countries, as well as the pooled data. In table 8, we find its effect on advertising likeability:
cognitive part is not significant in Russia and Japan, but significantly positive in other
countries and in the pooled data. In addition, in table 9 and 10, we find that the effects
of memory on both parts of advertising likeability are all positively significant (in most
case, p<0.001) and consistent, except for in Russia. Thus, we can safely conclude that
H2b is strongly supported.
Table 9 mediation effect of advertising likeability between advertising memory and purchasing intention
Country Total effect Memory →purchasing intention Direct effect Memory → purchasing intention Indirect effect
Memory →Ad liking →purchasing intention
Proportion of total effect that is
mediated Through LC Through LA Total indirect Australia
+
*+
-
*+
*+
* 94% Russia+
*+
-
+
+
51% China+
*+
-
*+
*+
* 99% Poland+
*+
-
*+
*+
* 93% Mexico+
*+
*-
*+
*+
* 64% Japan+
-
*-
*+
*+
* 337% Pooled without Japan+
*+
-
*+
*+
* 97% Pooled with Japan+
*+
-
*+
*+
* 98%1. “+” means the coefficient is greater than 0, and “-” means the coefficient is smaller than 0. 2. “+” or “-” which is with “*” means its 95% confidence intervals does not contain 0.
3. “Proportion of total effect that is mediated” equals to the ratio of “total indirect” to “total effect”. Table 10 mediation effect of advertising likeability between advertising memory and Word-of-Mouth
Country Total effect Memory → WOM
Direct effect Memory →WOM
Indirect effect
Memory →Ad liking →WOM Proportion of total effect that is
Japan
+
-
-
*+
*+
* 105% Pooled without Japan+
*+
-
*+
*+
* 93% Pooled with Japan+
*+
*-
*+
*+
* 85%1. “+” means the coefficient is greater than 0, and “-” means the coefficient is smaller than 0. 2. “+” or “-” which is with “*” means its 95% confidence intervals does not contain 0.
3. “Proportion of total effect that is mediated” equals to the ratio of “total indirect” to “total effect”.
In table 9 and 10 (for the detailed results, please refer to Appendix C), we find that
the indirect effects of advertising memory on two consumer behaviors through two parts
of advertising likeability are statistically different from 0, as revealed by 95%
bias-correlated bootstrap confidence intervals that is entirely above 0 (for advertising
likeability: affective part) or entirely below 0 (for advertising likeability: cognitive part)
except for in Russia. And we can see that the negative influence of advertising memory
on two consumer behaviors carried by advertising likeability: cognitive part is caused by
negative impact of advertising likeability: cognitive part on two consumer behaviors
(seeing Appendix C). We further find that proportion of total effect of adverting
memory on two consumer behaviors that is mediated is substantial, except for in Russia,
where the indirect effect does not account for 60% of the total effect. To sum up, we can
safely conclude that H2 is strongly supported.
Effect of advertising likeability on advertising memory
Table 11 effect of advertising likeability on advertising memory
DV: Ad memory (HLM) (1) AUS (2) RUS (3) CHI (4) POL (5) MEX (6) JAP (7) pooled without Jap (8) pooled with Jap LA 0.10*** 0.04 0.21*** 0.11*** 0.13*** 0.06** 0.12*** 0.11*** LA * LA -0.02 -0.01 0.06*** -0.02 0.03 0.04** 0.01 0.01 LC 0.18*** 0.07* 0.07 0.11*** 0.08* 0.02 0.12*** 0.10*** LC * LC 0.08*** 0.08*** 0.05* 0.10*** 0.04* 0.09*** 0.08*** 0.09***
For the estimation results of control variables, please refers to the appendix
N 3089 2402 2610 3768 3032 2713 14901 17614
Log restricted-likelihood -4233.22 -3173.35 -3581.86 -5211.93 -4149.46 -3719.39 -20239.21 -23609.85
* p < 0.05, ** p < 0.01, *** p < 0.001
20 / 59
the estimation results for the effect of advertising likeability on advertising memory. We
find that the quadratic terms of LC are significantly positive and consistent between
different countries, as well as the pooled data. The locations of LC for the lowest scores
of advertising memory are 2.73, 3.30, 4.05, 3.19, 3.33, 3.23, 3.12, and 3.12 for Australia,
Russia, China, Poland, Mexico, Japan, pooled data without Japan, and pooled data with
Japan respectively. Obviously, there is a “U-shape” relationship between advertising
likeability: cognitive part and advertising memory.
For LA, the empirical findings are relatively complicated and less consistent. In China
and Japan, quadratic terms of LA are significantly positive. And the locations of LA for
the lowest score of advertising memory are 2.41 and 2.54 for China and Japan
respectively. And in Mexico and pooled data with Japan, the t-values of the quadratic
term of LA are 1.96 and 1.92 respectively, which are marginal significant. In the rest
countries and the pooled data without Japan, except for Russia, the quadratic terms of
LA are not significant, but its monomial terms do. So, LA can affect advertising memory
through the linear approach in Australia, Poland, Mexico, and pooled data, or through
nonlinear approach in China and Japan. To sum up, H3a is weekly supported, and H3b is
substantially supported.
Robust checks
of advertising memory is used. However, for brand recall, the results are less consistent
with the results we obtained above: there is positive relationship between brand recall
and LA in four countries (Australia, Poland, Mexico, and Japan) and both pooled data
sets. This linear relationship became “U-shape” in China. The relationship between
brand recall and LC are quite diversified: positive linear in Australia, Russia, and Poland;
“U-shape” in Japan and both pooled data sets; “invert U-shape” in Mexico. Hence, it
seems that the properties of advertising memory in our analysis are heavily dominated by
message recall, not brand recall. Just as we mentioned before, brand recall just covers the
brand part of advertising memory, not the overall evaluation of advertising memory,
which may be the reason that there is not general relationship between advertising
memory and advertising likeability if advertising memory is measured by solo brand
recall.
Furthermore, we originally use OLS to estimate the equation 1 and 2. And in order to
account for the correlation between observations, we use HLM to estimate the equation
3, 4, and 5. In fact, just as we mentioned above, correlation mainly comes from the data
collection procedure: each respondent are asked to evaluate 8 pieces of commercials. So,
observations which come from the same respondents are correlated with each other.
OLS with special residual structure settings, which allows for this correlation, could
substantially deal with correlation
5. Thus, in robust checks procedure, we use OLS with
special residual structure settings to estimate the equation 3, 4, and 5, and use random
intercept model to estimate the equation 1 and 2. The comparison of estimation results
from different methods shows that the valance and the magnitude of key variables in
these five equations are quite similar.
In addition, we originally estimate these 5 equations separately, which means we
assume that the residuals of each equation is set to be independent with each other.
However, errors in each equation associated with dependent variables may be correlated
(Zellner, 1962, 1963; Zellner & Huang, 1962). Thus, in robust checks procedure, we relax
the assumption of the independence of errors between different equations, and
re-estimate these five equations simultaneously using seemingly unrelated regression. The
22 / 59
Breusch–Pagan test shows that the correlations of errors from different equations are
significant (Breusch & Pagan, 1980). The results of seemingly unrelated regression are
largely consistent with the results from the separated estimated models.
In the mediation test, we originally use the 95% bias-corrected bootstrap confidence
intervals to justify the indirect effect of two parts of advertising likeability. In the robust
checks procedure, we use the Percentile-based bootstrap confidence intervals and the
Monte Carlo confidence intervals. Both intervals can be obtained in PROCESS (A.F.
Hayes, 2013). Again, the results are quite consistent with their counterparts we originally
obtained.
Dual relationships between memory and likeability
In our conceptual model, we explicitly assume that advertising memory is an
antecedent of advertising likeability, and advertising likeability carries the influence of
advertising memory on purchasing intention and WOM. We further explicitly assume
that advertising likeability can also influence advertising memory. Both hypotheses are
substantially supported. Thus, there may be dual relationships between advertising
memory and advertising likeability. However, the data this paper is based on is
cross-sectional data, not time series data, which fundamentally limits our ability to figure
out which one is the first step. Although former literatures recommend that simultaneous
equation system with three-stage least squares can be applied to justify the
dual-relationships based on cross-sectional data(e.g. (Reibstein, Lovelock, & Dobson,
1980; Verhoef & Leeflang, 2009)), the simultaneous equation system we build cannot be
identified because of lacks of enough independent variables. Thus, we can justify the
dual relationships, not by statistical methods based on the data in this paper, but by
theatrical reasoning, as well as the significant effects revealed in the separately estimated
models.
6. General discussion
Word-of-Mouth. Hypotheses we proposed are all well supported except for the
relationship between likeability: cognitive part and two consumer behavior-oriented
measures. The empirical findings consistently reveal that the easier the commercial is to
understand, the less likely the consumer will buy the product or tell other people this
commercial. Here we provide some potential theoretical explanations. We originally
thought that if the commercial can be easily processed, consumers will easily perceive the
persuasive information contained in the commercial. And they will be more likely to
comply with the commercial and be persuaded to buy or to show the WOM behavior.
However, research in meta-cognition reveals that meta-cognitive beliefs about marketer’s
use of persuasion tactics and motives can be formed through effortful processes
(Friestad & Wright, 1995). And these beliefs usually lead to consumer skepticism, distrust,
and resistance to persuasion, especially when underlying motives of marketers is
obviously shown to the consumer (Fennis & Stroebe, 2010; Warlop & Alba, 2004). Thus,
if the commercial is easy to understand, its ulterior purpose can be easily perceived,
which will trigger consumer’s negative evaluation to the commercial, and finally, the
negative consumer behavior toward the commercial.
This paper explicitly reveals the impotent role of advertising likeability in determining
purchasing behavior, Word-of-Mouth, and in generating advertising memory, which
echoes the former studies which conclude advertising liking was the stronger factor
linked to persuasion and sales (Haley & Baldinger, 1991; Smit et al., 2006). Consumers
tend to buy the product and tell someone else this commercial if they like this
commercial. Commercial design may be an effective tactics to boost the sales.
However, not all advertising likeability parts have positive influence on purchasing
intention and Word-of-Mouth. If the commercial is straightforward and full of explicit
persuasion, consumer will generate the negative feelings toward the commercial, and
finally, decrease the intention of buying and Word-of-Mouth. Our findings here are quite
consistent with the meta-cognition studies (e.g. (Fennis & Stroebe, 2010; Warlop & Alba,
2004)). So, effective commercials entail the explicit information, as well as not irritating
the consumer by straightforward persuasion.
24 / 59
memory. For one thing, advertising likeability carries most of influence of advertising
memory on consumer behaviors. And the direct effects of memory are trivial. Memory
may be the necessary step of attitude formation. But consumer may be influenced by the
arguments even if they cannot retain them, or they may still remember the arguments,
even though they think they are invalid (Fennis & Stroebe, 2010). For another, consumer
will selectively retain the commercials which are quite good or quite bad. Liking or
disliking the commercials can both act as cues when consumers encode, store, and
retrieve the advertising information. Especially for advertising likeability: cognitive part,
consumer can easily retain the commercials which are easily processed. If the
commercials cannot be fully comprehended at the short time, consumers may pay much
attention to them, which may increase the extent to which the commercials can be
retained. This situation is more likely to occur in the forced exposure test, which we use
in the data collection procedure.
Another main purpose of this paper is to check to what extent the relationships
between memory, likeability, and consumer behavior are consistent between different
countries. The estimation results reveal that these relationships are quite robust between
different countries. Because culture difference is commonly used by social scientists to
represent the difference between countries, we can say that the effect of advertising
likeability is a kind of “culture free” effect, although cultural values are powerful forces
that shape perceptions and behaviors (Engelen & Brettel, 2011).
7. Limitations and future research
direct effect of implicit memory on consumer behaviors. Thus, the theoretical model and
the findings will be changed if implicit memory is used. The third limitation is that range
of sample collection can be expanded. Additional countries can be further included, such
as US, Germany, France, and Brazil. The third limitation is that when we examine the
effect of advertising memory on advertising liking, there is a censoring problem: we
cannot get the advertising liking evaluation from the respondents who do not have any
advertising memory. Consumers cannot infer the validity of information if they do not
have any memory about this information. So, the conclusion concerning the effect of
advertising memory on advertising liking is only applicable to the consumers who have
more or less information about the advertising, not to newcomer.
26 / 59
Appendix A: outcomes for the effect of advertising likeability (OLS and HLM)
Table 12 effect of advertising likeability on purchasing intention
DV: Purchasing intention (OLS) (1) Australia (2) Russia (3) China (4) Poland (5) Mexico (6) Japan (7) pooled without Jap (8) pooled with Jap Ad likeability: affective 0.794*** 0.795*** 0.830*** 0.778*** 0.759*** 0.669*** 0.801*** 0.787*** (50.30) (48.78) (51.05) (59.51) (51.40) (36.75) (123.78) (128.68) Ad memory 0.00655 0.0237 0.00173 0.00656 0.0387** -0.0313 0.00266 0.00201 (0.47) (1.63) (0.14) (0.51) (2.84) (-1.73) (0.45) (0.36) Ad likeability: cognitive -0.142*** -0.0634*** -0.0305 -0.133*** -0.0874*** -0.0676*** -0.0971*** -0.0905*** (-9.22) (-3.62) (-1.85) (-9.83) (-5.83) (-3.43) (-14.62) (-14.05) Product or service (1: product / 0:
service)
-0.127* -0.191* 0.0719 -0.0539 0.135 0.114 -0.0441 -0.0298
(-2.01) (-2.42) (1.55) (-0.76) (1.50) (1.72) (-1.86) (-1.45) Ad size (seconds) -0.00276*** 0.00275 -0.00187 -0.00441*** -0.0000691 -0.00182***
(-3.32) (1.33) (-1.59) (-3.56) (-0.09) (-4.75)
Gender (1: male /0: female) 0.0956** 0.0911* 0.0339 0.0226 -0.0240 -0.0429 0.0749*** 0.0630***
Sector: FMCG food 0.209** 0.248*** 0.0151 0.112 0.290* 0.140 0.195*** 0.178*** (2.86) (5.19) (0.31) (1.91) (2.33) (1.54) (6.85) (6.50) Sector: automobile -0.114 -0.0167 -0.198** 0.121 -0.271** -0.00250 -0.0425 (-1.38) (-0.32) (-3.24) (0.97) (-2.68) (-0.08) (-1.46) Sector: telecom 0.0174 -0.100 0.0337 -0.106 0.313** -0.169 0.00796 -0.00832 (0.33) (-1.26) (0.51) (-1.12) (3.12) (-1.55) (0.29) (-0.30) Sector: finance 0.0561 -0.0828 -0.0890 -0.117 0.0694 -0.0403 -0.0582* (1.15) (-1.09) (-1.18) (-1.30) (0.79) (-1.44) (-2.10) Sector: retailer 0.603*** 0.0195 0.115 0.0739 0.540*** 0.210*** 0.207*** (6.53) (0.20) (1.54) (0.83) (5.93) (6.06) (5.90) Sector: leisure 0.0357 0.334*** 0.247** 0.0156 0.0895* (0.53) (3.64) (2.91) (0.36) (2.51) Sector: utilities -0.102 -0.123 -0.129 (-1.25) (-1.66) (-1.71)
Sector: clothing & jewelry 0.155 -0.107 0.178** 0.0133
(1.93) (-1.08) (2.87) (0.28)
Sector: other (reference)
Education: Attended or completed some vocational / professional training
28 / 59
Education: Attended some college or got some degree
0.0501 -0.0613 0.205** -0.0528 -0.0916* 0.0221 0.0529** 0.0463**
(1.23) (-0.91) (2.91) (-1.54) (-2.14) (0.43) (2.79) (2.62)
Education: Less than high school or high school graduate
(reference)
Occupation: working in full or part time
0.0217 0.309** 0.131 0.0385 0.0232 0.0623 0.0869* 0.0886**
(0.32) (3.08) (0.70) (0.54) (0.38) (0.67) (2.45) (2.71) Occupation: study / education -0.0419 0.129 -0.241 0.0153 0.0111 0.0804 -0.0276 -0.0208
(-0.46) (0.98) (-0.85) (0.19) (0.12) (0.54) (-0.60) (-0.48) Occupation: retired, or disabled, or
housewife / man
-0.0438 0.220* 0.241 0.0462 -0.0395 -0.0807 -0.0152 -0.0180
(-0.63) (2.03) (1.11) (0.58) (-0.52) (-0.75) (-0.39) (-0.50)
Occupation: unemployed (reference)
Household situation: Living with (grand)parents / relatives
-0.253** 0.141 -0.189 0.00732 0.000521 -0.0417 -0.0522 -0.0681
(-2.73) (1.05) (-1.15) (0.09) (0.01) (-0.40) (-1.17) (-1.67) Household situation: Single, no child
or with child(ren)
-0.0548 0.0594 -0.132 0.0711 0.0268 -0.0591 -0.0111 -0.0285 (-0.71) (0.46) (-0.77) (0.90) (0.29) (-0.58) (-0.26) (-0.73) Household situation: Student house,
commune
-0.219 0.548** 0.151 -0.0412 -0.770*** -0.0109 -0.0179
(-1.27) (3.09) (0.57) (-0.37) (-5.49) (-0.15) (-0.24) Household situation: Married /
living together
Household situation: other (reference) Constant 0.0881 -0.458* -0.223 0.144 -0.308* -0.114 -0.155** -0.185*** (0.83) (-2.53) (-0.89) (1.05) (-2.25) (-0.80) (-2.67) (-3.48) N 3089 2402 2610 3768 3032 2713 14901 17614 adj. R2 0.544 0.600 0.677 0.550 0.554 0.416 0.603 0.589 Max VIF 5.88 10.05 19.32 6.91 23.15 8.31 6.16 6.43 Mean VIF 2.53 3.54 4.82 2.97 5.09 3.27 2.63 2.72 Min VIF 1.06 1.08 1.04 1.05 1.03 1.04 1.02 1.04 t statistics in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001
Table 13 effect of advertising likeability on WOM
DV: WOM (OLS) (1) Australia (2) Russia (3) China (4) Poland (5) Mexico (6) Japan (7) pooled without Jap (8) pooled with Jap Ad likeability: affective 0.651*** 0.718*** 0.798*** 0.689*** 0.672*** 0.509*** 0.716*** 0.699*** (23.99) (35.95) (42.13) (38.55) (41.07) (22.09) (87.50) (88.66) Ad memory -0.00107 0.0321 -0.0104 0.0498*** 0.0101 -0.00223 0.00602 0.0149* (-0.06) (1.86) (-0.71) (3.35) (0.63) (-0.11) (0.85) (2.20) Ad likeability: cognitive -0.174*** -0.0585** -0.0643*** -0.135*** -0.0506** -0.0510* -0.0972*** -0.0835*** (-8.41) (-2.97) (-3.42) (-8.60) (-3.24) (-2.45) (-12.94) (-11.34) Product or service (1: product / 0:
service)
-0.0155 -0.0168 0.00570 -0.0294 -0.0218 -0.00925 -0.0822** -0.0731**
(-0.22) (-0.19) (0.11) (-0.35) (-0.25) (-0.11) (-3.13) (-3.18) Ad size (seconds) 0.000658 0.00666** -0.00209 0.00345** -0.000708 0.00134**
30 / 59
Gender (1: male /0: female) 0.0529 0.131** 0.0544 -0.0969* 0.00466 0.00367 0.0693*** 0.0679***
(-1.61) (-3.34) (-3.08)
Sector: clothing & jewelry -0.179 0.0647 -0.00226 -0.0472
(-1.83) (0.59) (-0.03) (-0.93) Sector: other (reference)
Education: Attended or completed some vocational / professional training
-0.0418 0.0793 0.0156 -0.00445 -0.0390 -0.000180 0.0400 0.0506 (-0.61) (0.84) (0.11) (-0.04) (-0.64) (-0.00) (1.28) (1.70) Education: Attended some college or
got some degree
0.0811 0.0425 0.0526 -0.0829 -0.110 0.0716 0.0491 0.0457 (1.26) (0.48) (0.53) (-1.80) (-1.74) (1.16) (1.87) (1.90) Education: Less than high school or
high school graduate (reference)
Occupation: working in full or part time -0.0458 0.401** 0.0795 0.0196 0.0292 0.307** 0.0913 0.120**
(-0.40) (2.67) (0.54) (0.24) (0.33) (2.62) (1.86) (2.64) Occupation: study / education -0.199 0.213 -0.389 -0.0339 -0.212 0.245 -0.141* -0.0973
(-1.29) (1.18) (-1.20) (-0.33) (-1.75) (1.16) (-2.25) (-1.64) Occupation: retired, or disabled, or
housewife / man
-0.0630 0.378* 0.0115 0.0126 -0.108 0.123 -0.0479 -0.0255
(-0.54) (2.39) (0.05) (0.13) (-0.97) (0.91) (-0.88) (-0.50) Occupation: unemployed (reference)
Household situation: Living with (grand)parents / relatives
-0.371** 0.191 -0.305 -0.0254 -0.104 0.0774 -0.101 -0.121*
32 / 59
Household situation: Single, no child or with child(ren)
-0.217 0.136 -0.136 0.00787 -0.101 0.0765 -0.0946 -0.0997 (-1.94) (0.81) (-0.80) (0.08) (-0.78) (0.52) (-1.69) (-1.89) Household situation: Student house,
commune
-0.371 0.127 -0.167 -0.132 -0.554* -0.201* -0.215*
(-1.51) (0.51) (-0.49) (-0.83) (-2.01) (-2.18) (-2.34) Household situation: Married / living
together
-0.159 0.153 -0.239 -0.0165 -0.00439 0.135 -0.0224 -0.0190 (-1.51) (0.97) (-1.60) (-0.18) (-0.03) (0.98) (-0.43) (-0.38) Household situation: other (reference)
Constant 0.250 -0.906*** 0.173 0.202 -0.0449 -0.412* -0.142 -0.0849 (1.57) (-3.88) (0.76) (1.31) (-0.25) (-2.19) (-1.86) (-1.20) N 3089 2402 2610 3768 3032 2713 14901 17614 adj. R2 0.336 0.484 0.597 0.425 0.446 0.253 0.490 0.479 Log restricted-likelihood Max VIF 5.88 10.05 19.32 6.91 23.15 8.31 6.16 6.43 Mean VIF 2.53 3.54 4.82 2.97 5.09 3.27 2.63 2.72 Min VIF 1.06 1.08 1.04 1.05 1.03 1.04 1.02 1.04 t statistics in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001
Table 14 effect of advertising likeability on purchasing intention
Ad memory 0.0000941 0.0237 0.00225 0.00489 0.0456*** -0.0310* 0.00967 0.00486
(0.01) (1.67) (0.19) (0.43) (3.63) (-2.02) (1.81) (0.95) Ad likeability: cognitive -0.146*** -0.0599*** -0.0305* -0.134*** -0.0885*** -0.0680*** -0.0963*** -0.0942***
(-10.13) (-3.72) (-2.28) (-10.62) (-6.51) (-3.92) (-16.10) (-16.22) Product or service (1: product / 0: service) -0.123 -0.190 0.0713 -0.0541 0.140 0.115 -0.0599 -0.0555 (-0.98) (-1.51) (1.28) (-0.57) (0.91) (1.18) (-0.93) (-0.99) Ad size (seconds) -0.00306 0.00322 -0.00185 -0.00452** -0.000170 -0.00222*
(-1.67) (0.85) (-1.36) (-3.09) (-0.12) (-2.26)
Gender (1: male /0: female) 0.0982*** 0.0908*** 0.0333 0.0229 -0.0219 -0.0436 0.0484*** 0.0394***
34 / 59 (0.04) (-0.81) (0.47) (-0.86) (1.71) (-0.46) (0.22) (-0.03) Sector: finance 0.0497 -0.0801 -0.0900 -0.120 0.0651 0.104 -0.0322 -0.0589 (0.46) (-0.60) (-1.24) (-0.96) (0.43) (0.73) (-0.42) (-0.82) Sector: retailer 0.615** 0.0277 0.115 0.0695 0.538** 0.223* 0.225* (3.18) (0.17) (1.23) (0.61) (2.97) (2.25) (2.27) Sector: leisure 0.0528 0.343* 0.347** 0.0272 0.0760 (0.41) (2.16) (2.79) (0.25) (0.81) Sector: utilities -0.102 -0.139 -0.152 (-0.60) (-0.71) (-0.78)
Sector: clothing & jewelry 0.152 0.188 0.0183
(1.37) (0.96) (0.15)
Sector: other (reference)
Education: Attended or completed some vocational / professional training
-0.0187 -0.00874 0.0463 -0.0568 -0.0659 -0.0290 -0.0263 -0.0262 (-0.57) (-0.16) (0.58) (-0.86) (-1.87) (-0.56) (-1.56) (-1.64) Education: Attended some college or got some
degree
0.0517 -0.0611 0.204*** -0.0518* -0.0920* 0.0232 -0.00677 -0.00378
(1.76) (-1.20) (3.70) (-2.18) (-2.54) (0.64) (-0.50) (-0.30)
Education: Less than high school or high school graduate (reference)
Occupation: working in full or part time 0.0199 0.318*** 0.139 0.0412 0.0130 0.0596 0.0556* 0.0576*
Occupation: study / education -0.0405 0.143 -0.229 0.0186 0.0105 0.0713 -0.0151 -0.0121 (-0.52) (1.36) (-1.02) (0.29) (0.15) (0.58) (-0.44) (-0.37) Occupation: retired, or disabled, or housewife /
man
-0.0399 0.223* 0.249 0.0474 -0.0504 -0.0825 -0.00869 -0.0141
(-0.70) (2.37) (1.11) (0.81) (-0.79) (-1.00) (-0.30) (-0.52)
Occupation: unemployed (reference)
Household situation: Living with (grand)parents / relatives
-0.251*** 0.144 -0.187 0.0105 -0.00105 -0.0387 -0.0483 -0.0464
(-3.74) (1.43) (-1.43) (0.19) (-0.02) (-0.43) (-1.63) (-1.67) Household situation: Single, no child or with
child(ren)
-0.0582 0.0593 -0.130 0.0754 0.0281 -0.0547 0.000349 -0.00823 (-1.07) (0.62) (-0.98) (1.42) (0.42) (-0.61) (0.01) (-0.31) Household situation: Student house, commune -0.238 0.549*** 0.148 -0.0383 -0.740*** -0.0186 -0.0243
(-1.78) (3.95) (0.82) (-0.44) (-3.83) (-0.37) (-0.48) Household situation: Married / living together -0.0427 0.0690 -0.120 0.0581 0.0949 0.0145 0.0230 0.0222 (-0.84) (0.76) (-0.96) (1.23) (1.51) (0.17) (0.89) (0.90)
Household situation: other (reference)
36 / 59
(-32.52) (-31.71) (-41.22) (-34.44) (-33.39) (-19.69) (-85.71) (-88.54)
N 3089 2402 2610 3768 3032 2713 14901 17614
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
Table 15 effect of advertising likeability on WOM
DV: WOM (HLM) (1) Australia (2) Russia (3) China (4) Poland (5) Mexico (6) Japan (7) pooled without Jap (8) pooled with Jap Fixed effect Ad likeability: affective 0.654*** 0.718*** 0.798*** 0.692*** 0.690*** 0.509*** 0.680*** 0.662*** (36.37) (40.96) (50.98) (49.27) (42.90) (26.44) (97.01) (101.14) Ad memory -0.00148 0.0342* -0.00908 0.0499*** 0.0164 0.00166 0.0161** 0.0178** (-0.09) (2.13) (-0.70) (3.86) (1.16) (0.10) (2.66) (3.10) Ad likeability: cognitive -0.174*** -0.0590** -0.0625*** -0.136*** -0.0561*** -0.0505** -0.0940*** -0.0912*** (-9.67) (-3.21) (-4.18) (-9.56) (-3.68) (-2.62) (-13.89) (-13.96) Product or service (1: product / 0: service) -0.0180 -0.0136 0.00650 -0.0277 -0.0180 -0.00882 -0.110 -0.101
(-0.20) (-0.11) (0.10) (-0.19) (-0.11) (-0.10) (-1.31) (-1.36) Ad size (seconds) 0.000714 0.00697 -0.00216 0.00348 -0.000742 0.00111
(0.55) (1.82) (-1.33) (1.57) (-0.49) (0.88)
Gender (1: male /0: female) 0.0518 0.136*** 0.0534* -0.0919*** 0.00701 0.00392 0.0273* 0.0288**
(1.64) (4.31) (2.01) (-3.63) (0.25) (0.11) (2.34) (2.62) Age (centered) -0.0569*** -0.0197* -0.0138 0.00540 -0.0272*** -0.0423*** -0.0212*** -0.0235***
(-0.64) (-0.16) (-0.57) (-2.10) (0.88) (0.03) (1.09) (0.80) Sector: FMCG Nonfood -0.140 0.0720 0.0628 -0.145 -0.0377 -0.0373 0.0196 -0.0100 (-1.32) (0.89) (0.60) (-1.25) (-0.16) (-0.38) (0.19) (-0.10) Sector: FMCG food -0.0197 0.0747 -0.00182 -0.131 0.239 -0.0944 0.234* 0.178 (-0.18) (0.99) (-0.03) (-1.14) (1.03) (-1.05) (2.40) (1.89) Sector: automobile -0.0897 0.0454 -0.274* 0.137 -0.132 0.147 0.102 (-0.75) (0.59) (-2.26) (0.59) (-1.32) (1.40) (1.01) Sector: telecom -0.0776 0.0348 -0.0982 -0.382* 0.150 -0.168 -0.0625 -0.0665 (-0.99) (0.25) (-1.16) (-1.99) (0.75) (-1.28) (-0.64) (-0.68) Sector: finance -0.0143 0.118 -0.150 -0.193 0.185 -0.0661 0.0227 -0.0288 (-0.18) (0.88) (-1.72) (-1.02) (1.11) (-0.51) (0.23) (-0.30) Sector: retailer -0.0279 0.0149 -0.133 -0.299 0.233 0.0310 0.0331 (-0.20) (0.09) (-1.18) (-1.74) (1.17) (0.24) (0.25) Sector: leisure 0.0384 0.186 -0.00128 0.0113 -0.0152 (0.41) (1.06) (-0.01) (0.08) (-0.12) Sector: utilities -0.164 -0.296 -0.288 (-1.33) (-1.17) (-1.11)
Sector: clothing & jewelry -0.184 0.0133 -0.0443
(-1.09) (0.05) (-0.27)
