Memory in Online Video Advertising:
Investigating the Viewing Process
Effects of device and viewing time on the relationship between
emotional tone and involvement on memory
Jessica Attema
S3139743
j.attema.1@student.rug.nl
MSc Marketing Intelligence
University of Groningen
Faculty of Economics and Business
Supervisor: Prof. Dr. T.H.A. Bijmolt
Co-assessor: Dr. P.S. van Eck
PREFACE
Four and a half years ago, I started my time as a student in Groningen. During my bachelor International Business I became increasingly interested in marketing, and specifically data analytics. I decided to pursue this interest by following the master Marketing Intelligence. Now, at the end of my student life, I look back at an incredible time. I have greatly enjoyed the past 4,5 years and I am excited to see what the future holds.
I would like to thank my supervisor Prof. Dr. T.H.A. Bijmolt for his guidance and feedback I received while writing this thesis. I would also like to thank Mark and Lisette from DVJ Insights for providing me with the data that is used in this study.
I hope you will enjoy reading my thesis.
ABSTRACT
Online video advertising is an increasingly popular form of digital advertising. However, its effectiveness is not well understood. This paper sheds light on this issue by investigating which factors affect participants’ memory of online video advertisements. Building upon the Limited Capacity Model of Motivated Mediated Message Processing and The Elaboration Likelihood Model, this paper posits the emotional tone of the advertisements and participants’ involvement with the advertisement as antecedents of memory. Next to that, effects of viewing time and participants’ device are analyzed.
Data was obtained from a market research agency from the Netherlands. An online questionnaire was conducted where participants from three countries, the Netherlands, Germany and the United Kingdom, viewed two skippable, online video advertisements on a video platform. After finishing, their emotional tone and level of involvement were measured, as well as their advertising memory.
The results indicate that the emotional tone and involvement positively impact participants’ memory. Viewing time is found to partially mediate these relationships. Furthermore, this paper finds that advertising memory is higher for participants viewing the video advertisement on a desktop or laptop rather than on a mobile device. The device category also moderates the partial mediation of viewing time, with the relationship being stronger for desktops and laptops. These findings are important for marketers and brand managers as it helps understand which are important elements to focus on when creating online video advertisements. Furthermore, advertising budgets can be allocated more efficiently between device categories based on the findings of this paper.
Keywords: online video advertising, emotional tone, involvement, viewing time, device
TABLE OF CONTENT
1. INTRODUCTION ... 4
2. THEORETICAL BACKGROUND AND HYPOTHESES ... 6
2.1 Emotional Tone ... 6
2.2 Involvement ... 8
2.3 Viewing Time ... 9
2.4 Device Category ... 11
3. METHODOLOGY ... 13
3.1 Sample and Data Collection ... 13
3.2 Procedure ... 13
3.3 Measures ... 14
4. DATA ANALYSIS ... 17
4.1 Pre-analysis Examination and Descriptive Statistics ... 17
4.2 Data Transformations ... 18
5. RESULTS ... 19
5.1 Memory – Emotional Tone ... 19
5.2 Memory – Involvement ... 21
5.3 Ad Recall – Emotional Tone ... 22
5.4 Ad Recall – Involvement ... 24 5.5 Model Fit ... 25 5.6 Sensitivity Analysis ... 26 5.7 Predictive Validity ... 27 5.8 Control Variables ... 28 6. DISCUSSION ... 30 6.1 Theoretical Implications ... 30 6.2 Practical Implications ... 31
6.3 Limitations and Further Research ... 32
7. REFERENCES ... 34
APPENDIX A: SURVEY ... 38
APPENDIX B: SMARTPLS ... 41
APPENDIX C: COUNTRY RESULTS ... 43
1. INTRODUCTION
Digital video advertising is a continuously growing form of digital advertising. Globally, in 2019, advertisers spent on average $17.8 million on digital video advertising, 25 percent more than the year before (IAB 2020a). Out of the several forms of digital advertising, video advertising is the fastest growing form, with revenues increasing by 33.5% in 2019 over 2018, amounting to $21.7 billion globally (IAB 2020b). Video advertising on online platforms such as YouTube is the most prevalent, constituting 60 percent of digital video advertising expenditure (IAB 2020a). This increasing popularity of online video advertising calls for a greater understanding for marketers of the drivers of the effectiveness of these ads.
Online video advertisements placed on platforms such as YouTube can mostly be skipped after a certain amount of time has elapsed. To make an impact, it is therefore of crucial importance to capture a viewer’s attention in these first seconds of exposure. Despite the wide availability of literature on online video advertising, it is not well understood which aspects of the video advertisements and the viewing process capture the viewers’ attention.
Divergent metrics have been used to measure advertising effectiveness. These include attitude towards the ad or the advertised brand (Belanche, Flavián and Pérez-Rueda 2017; Stewart et al. 2019), purchase intention (Fortin and Dholakia 2005; Stewart et al. 2019), intention to share the advertisement (Eckler and Bolls 2011; Tellis et al. 2019) and advertising memory (Li and Lo 2015; Song, Chan and Wu 2019). It is the latter measure of advertising effectiveness this research will focus on. It is of crucial importance to marketers and brand managers as it is related to brand awareness. A higher brand awareness increases the possibility of consumers considering the brand when making a purchase decision. In addition, brand awareness is a necessary condition for creating lasting associations in the consumers’ hearts and minds, which together constitute the building block towards customer-based brand equity. Understanding what influences this brand equity can ultimately improve marketing productivity (Keller 1993). Therefore, knowing what influences the memory of ads and advertised brands is of crucial importance to marketers and brand managers.
the advertisement. Perceived involvement measures ‘a person’s perceived relevance of the object based on inherent needs, values, and interests’ (Zaichowsky 1985), while emotional tone measures the degree to which a response to a stimulus is perceived as positive or negative (Lang 1979). This paper therefore examines viewing time and memory in combination with emotional tone and involvement.
Related to the viewing process, is the device used to display the video advertisement. This is a relevant question for managers in determining how to allocate advertising budgets. Limited research is available examining the effect of the device on advertising effectiveness. Some research including device category has examined its influence on attitudes toward the ad or the advertised brand (Stewart et al. 2019) but the effect of the device category on memory of ads and advertised brands has not been studied before. This calls for a greater understanding of how the device used to view an advertisement affects memory. Therefore, this paper contributes to the literature by taking into account the device category in measuring advertising memory.
While most studies have focused solely on aided recall (brand recognition) or unaided recall (brand recall) (Li and Lo 2015; Song, Chan and Wu 2019), this paper extends current research on advertising memory by incorporating two other measures of memory; namely advertisement and message recall. This is relevant for marketers and brand managers as this reveals the specific impact of the message, rather than a mere focus on the impact of the brand.
This paper contributes to the literature by (1) using memory as a measure of advertising effectiveness, (2) taking into account the viewing time as a mediator, (3) taking into account the device used when watching online video advertisements and (4) using several measures of memory. This is relevant for marketers and brand managers as a higher memory leads to a higher brand awareness, which increases brand equity and marketing productivity. Furthermore, these insights can be used to better allocate advertising budgets.
This research aims to answer the question: How do the emotional tone of the ad and
involvement with the ad influence advertising memory and how do the viewing time and the device category affect this relationship? This paper aims to answer this question by examining
2. THEORETICAL BACKGROUND AND HYPOTHESES
This chapter will provide a theoretical background for the proposed relationships. To explain the influence of the emotional tone and involvement, The Limited Capacity Model of Motivated Mediated Message Processing and the Elaboration Likelihood Model are used respectively. Next to that, a theoretical reasoning for the mediator and moderator are provided. A summary of existing literature relevant to this paper is presented in Table 1, thereby identifying the research gap this paper aims to fill.
Paper Involvement Emotional Tone Viewing Time Device Effectiveness measure Results Li and Lo (2015) ✓ (Ad length) Brand Recognition Ad length positively influences brand recognition Belanche, Flavián and Pérez-Rueda (2017) ✓ ✓ (Arousal) ✓ Attitude, Viewing Time Arousal increases viewing time Involvement not significant Song, Chan and Wu (2019) ✓ Brand Recognition Valence positively influences brand recognition. Arousal was not significant.
Stewart et al. (2019)
✓ ✓ Ad/Brand
Attitude, PI, Seeking Info, Opt-in for Info
Involvement influences PI and opt-in for info. Device affects seeking info and opt-in for info
This research ✓ ✓ ✓ ✓ Advertising
Memory
Table 1. Overview of Past Research on Online Video Advertising
2.1 Emotional Tone
The emotional tone of advertisements is defined as the degree to which an emotional response to a stimulus is perceived as positive or negative (Lang 1979). It has often been identified as an important antecedent of advertising effectiveness (Teixeira, Wedel and Pieters 2012; Tellis et al. 2019). While most studies have focused on emotional tone in relation with attitudes toward the advertised brand or the advertisement itself (Eckler and Bolls 2011), some have extended this to memory. Lang and Dillon (1995) and Moorman, Nijens and Smit (2007) have found that the emotional tone of advertisements has a significant impact on cognitive processing and memory. Although the effect of emotional tone on advertising memory is often investigated for traditional forms of advertising, the current research will extend this to online video advertising.
tone of media messages affects the processing of these messages in the human mind. Specifically, the LC4MP provides theoretical explanations for how media messages capture attention, how they are processed and subsequently stored and recalled. It applies cognitive and motivational mechanisms to understand how the emotional tone of messages affects cognitive processing.
The LC4MP proposes that messages either activate the appetitive (approach) motivational system or the aversive (avoid) motivational system. The system which is activated depends on the motivationally relevant stimulus. As people inherently tend to maximize pleasantness (Reeves et al. 1991) different motivational reactions can be expected for advertisements perceived as pleasant (e.g. positive) versus unpleasant (e.g. negative) (Lang 2009). Perception of the stimulus, whether positive or negative, is largely relative to the different life experiences of the individual (Reeves et al. 1991).
According to the LC4MP, positively perceived stimuli activate the appetitive system, whereas negatively perceived stimuli activate the aversive system. The activation of each of these systems has consequences for the processing of the media messages. In line with the dual-system theory proposed by Cacioppo, Gartner and Berntson (1997), more cognitive resources will be allocated to messages activating the appetitive system. In contrast, fewer cognitive resources will be allocated to messages activating the aversive system. As the level of cognitive resources is directly related to the storage and retrieval of media messages, positive messages will be stored more effectively and can be retrieved more easily (Lang 2009; Eckler and Bolls 2011).
Several studies have examined the relationship between emotional tone and memory of brands and advertisements. Bolls, Lang and Potter (2001) have found that positive radio advertisements were better recognized and recalled than negative radio advertisements. Song, Chan and Wu (2019) found that people who reported positive valence were more likely to recognize brands in a video clip selected from a film than participants who reported a negative valence. As the LC4MP model can be applied to any type of media and any type of message (Lang 2009), this paper proposes that these results extend to the context of online video advertisements. Put formally;
H1: Advertising memory is higher for consumers who perceive the emotional tone of
2.2 Involvement
Involvement is defined as ‘a person’s perceived relevance of the object based on inherent needs, values, and interests’ (Zaichowsky 1985). Previous research has made a connection between the concept of involvement and its relationship with advertising. For example, Wright (1973) defined involvement with advertising as ‘the receiver’s perception of the relevancy of the ad content to some pending problem’. Involvement has therefore widely been acknowledged as an important factor explaining advertising outcomes.
Involvement is often measured from either of two differing perspectives. Some research measures involvement from a product’s perspective, suggesting that individuals are either high or low involved with a product (Stewart et al. 2019) They suggest that involvement depends on the product itself, where certain categories of products are always defined as either high or low involvement products. Conversely, a second stream of literature measures involvement from an individual’s perspective, where the level of involvement depends on the individual (Zaichowsky 1994; Belanche, Flavián and Pérez-Rueda 2017). This stream of literature recognizes that a product category does not always convey the same relevance to a person, and thus, that different individuals can constitute either high or low involvement with the same product. This latter stream of research projects is in line with the definitions of Zaichowsky (1985) and Wright (1973), which both focus on the individual when determining the level of involvement, rather than involvement being a static concept based on the product category. Even though some product categories tend to be inherently low or high involving (Zaichowsky 1985) this does not hold for all types of products (Zaichowsky 1994). Thus, this highlights the importance of measuring involvement with an advertisement as a person’s perceived relevance of the advertisement, which is the perspective this research follows.
persuasion, where aspects such as attractiveness or music are focused on. Thus, the content of the message will be evaluated less effectively (Petty and Cacioppo, 1981; Petty, Cacioppo and Schumann 1983).
When thinking of involvement in advertising, it can be said that consumers who are more involved pay more attention to the ad and its message and focus more on brand processing (Buchholz and Smith 1991). They are more motivated and exert a higher level of cognitive processing than low involved individuals. Specifically, this means that they process the information presented in the advertisement at a deeper level. As this processing becomes more sophisticated, recognition and memory of the advertisement are enhanced. In contrast, low involvement individuals process the advertisement at a more superficial level, focusing on attractiveness or auditory elements. This leads to fewer enduring relationships with the ad in terms of memory of the advertisement (Petty, Cacioppo and Schumann 1983; Buchholz and Smith 1991).
Several studies have investigated this proposed relationship. Buchholz and Smith (1991) have found that consumers with high involvement are better able to recognize brands and advertisements in radio and TV commercials. Similarly, Peltier and Schibrowsky (1994) found that highly involved individuals were better able to recall the advertised product presented in a slide show but found no effect for message recall. Even though these papers have focused on traditional forms of advertising, this paper proposes that the same effect can be expected in video advertising. Put formally;
H2: Advertising memory is higher for consumers who are highly involved with the
online video advertisement compared to consumers who are less involved.
2.3 Viewing Time
Several studies have examined the relationship between an advertisement’s length and the subsequent advertising memory (Allan 2007; Li and Lo 2015). These studies generally agree that when an advertisement is longer, the level of advertising memory is increased, as viewers have more time to attend to the stimulus and to effectively evaluate this stimulus. Although these studies have mostly focused on traditional forms of advertising such as television and radio ads (Allan 2007), it is expected that the same holds for online video advertisements.
that an advertisement can elicit in individuals might result in that individual to continue watching the advertisement for a longer period of time. This is especially relevant when video advertisements can be skipped, which is oftentimes the case on video platforms such as YouTube. A longer viewing time increases the likelihood of capturing attention, which can then result in higher levels of memory of the advertisements and advertised brands.
Belanche, Flavián and Pérez-Rueda (2017) found that video advertisements that elicited a high feeling of arousal increased the amount of time participants viewed the video. This study was done in the context of skippable video advertisements, and thus highlights that arousal can enhance advertising effectiveness. As arousal is one dimension associated with the emotional tone of advertisements (Lang 1979), one can expect that a more positive emotional tone increases viewing time. In addition, Olney, Holbrook and Batra (1991) found a positive relationship between the emotional appeal of advertisements, measured by pleasure and arousal, on viewing time. On the other hand, several studies have found a connection between the length of advertisements and memory of these advertisements. In the context of video advertising, Li and Lo (2015) have found that a greater ad length has a positive impact on recognition and recall. Putting the findings of Belanche, Flavián and Pérez-Rueda (2017) and Li and Lo (2015) together, it can be expected that a more positive emotional tone increases viewing time, which in turn increases memory of ads and advertised brands. Specifically;
H3: Viewing time mediates the relationship between emotional tone and memory of ads
and advertised brands, as emotional tone positively affects viewing time, which in turn positively affects memory.
that viewing time mediates the relationship between involvement and advertising memory. Put formally;
H4: Viewing time mediates the relationship between involvement and memory of ads
and advertised brands, as involvement positively affects viewing time, which in turn positively affects memory.
2.4 Device Category
Little research has devoted attention to the role of the device in the effectiveness of online video advertisements. Mobile devices are increasingly used when watching videos, and now constitute 68 percent of global digital video revenues in 2019, up from 52 percent in 2017 (IAB 2020b). This has led to researchers focusing on the effectiveness of mobile advertisements (Goh, Chu and Wu 2015; Grewal et al. 2016). However, this stream of literature only focuses on mobile devices and does not take into account possible differences in advertising effectiveness across device categories. This calls for a greater understanding of the effectiveness of online video advertising using different devices. This research adds to the literature by incorporating the effect of the device category on advertising effectiveness.
Viewing devices can be classified based on their differing screen sizes. Mobile devices such as smartphones and tablets have smaller screen sizes, whereas non-mobile devices such as desktops and laptops have larger screen sizes. Due to smaller screen sizes on mobile devices, an advertisement will take over a larger part of the screen, capturing more of the viewer’s attention (Grewal et al. 2016). Furthermore, interaction on mobile devices is touch-based, which differs from mouse-based interactions on computers. Particularly, Brasel and Gips (2014) propose that touch-based interactions can elicit ownership effects that could enhance advertising effectiveness. These findings suggest that viewing video advertisements on a mobile device should enhance advertising effectiveness.
shed light on this potential interaction. In line with the reasoning of Grewal et al. (2016) and Brasel and Gips (2014), this paper proposes that viewing online video advertisements on mobile devices increases memory of the ads and advertised brands.
H5: Mobile device users have a higher memory of online video advertisements and
advertised brands than do non-mobile device users.
Apart from the direct effect of the device category on advertising outcomes, this paper also proposes the device category as a moderator. Specifically, it is expected that the device category moderates both indirect effects of the mediation relationships proposed before. Following the reasoning of Grewal et al. (2016), Brasel and Gips (2014) and the finding by Joa, Kim and Ha (2018), this paper proposes that viewing online video advertisements on mobile devices strengthens the mediating relationship between emotional tone, involvement, viewing time and memory of ads and advertised brands.
H6: Device category moderates the mediation relationship between emotional tone,
viewing time and memory in such a way that the mediation relationship is stronger for users viewing the online video advertisement on a mobile device.
H7: Device category moderates the mediation relationship between involvement,
viewing time and memory in such a way that the mediation relationship is stronger for users viewing the online video advertisement on a mobile device.
The proposed hypotheses are summarized in the conceptual model in Figure 1.
3. METHODOLOGY
3.1 Sample and Data Collection
The data was obtained from a market research agency from the Netherlands. Participants were recruited from a panel in exchange for money or points that could be collected. The survey was available online to participants from the panel in The Netherlands, Germany and the United Kingdom. Together, the participants evaluated 94 different video advertisements: 32 German ads, 30 Dutch ads and 32 English ads. The advertisements are from companies or brands from each of the three countries. These include a wide variety of industries. Examples are food brands, supermarkets, insurance companies, electronic appliances and personal care products. As each of the 4846 respondents viewed 2 video advertisements, each ad was viewed approximately 100 times.
In total, 1598 (33%) participants from The Netherlands answered the survey, 1655 (34%) from Germany and 1593 (33%) from the United Kingdom. One response was dropped to due incomplete data, leading to a total of 9691 observations. Of all respondents, 50.8% were female and 49.2% male. The questionnaire was answered by participants between 18 and 90 years of age, with the average being 45 years.
3.2 Procedure
Participants could answer the survey on their chosen device. They were first asked to indicate their gender and age. Participants viewed six different web pages and were told that they could scroll freely through each web page, and could continue to the next web page when they wished. Each participant viewed four news websites with each one banner advertisement, along with two video platform pages with each one video advertisement. On the pages with a video advertisement, the video advertisement played for at least five seconds before the participants could skip it to watch the video. As the interest of this research is in the video advertisements, the banner ads merely functioned as a distraction so that participants did not presume the interest of the study.
after five seconds. They were asked to answer a number of statements about the advertisements, including the statements measuring the involvement and emotional tone variables. Finally, participants were thanked for their participation. Examples of survey questions are presented in Appendix A.
3.3 Measures
Dependent variable measures
Unaided recall (Brand recall) measures if the participants are spontaneously able to
recall the advertised brand. It is measured by an open-ended question asking the participants which brand they had just seen. If the correct brand is answered, this variable will receive the value ‘1’, and ‘0’ otherwise. Aided recall was measured by providing a list of six brand names, including to correct brand name. In addition, six competing brand names were listed. The brand names were presented in random order to rule out order effects. Participants were asked, for each brand, whether they had seen the brand on the website they browsed through. Answer options includes ‘Yes, definitely’, ‘Maybe’ and ‘No, definitely not’. If the respondent answered ‘Yes, definitely’ to the brand name of which he or she saw the video advertisement, this variable receives the value ‘1’, and ‘0’ otherwise. Message recall is measured by asking if the participants could remember the message the ads were trying to convey. Only the ad or ads that the participant indicated they recognized in the aided recall question were provided. This way, it is assumed that if participants did not recognize the ad using cues, the participant will also not recall the message the ad was trying to convey. This variable takes the value ‘1’ if the participants indicates ‘Yes’, and ‘0’ if they indicated ‘No’. Ad recall is measured by asking participants whether they have seen the advertisement, when they are presented with screenshots of the video. This variable takes the value ‘1’ when participants indicate ‘Yes, definitely’ or ‘Maybe’, and ‘0’ if they indicate ‘No’.
To measure participants’ level of advertising memory, the four separate recall items will be summed into one overall memory variable. Factor analysis and reliability analysis are performed to examine whether this is justified. KMO measure of sampling adequacy (.67) and Bartlett’s test of sphericity (p < .001) show that factor analysis is appropriate. However, the results of the factor and reliability analyses show that ad recall cannot be summed with the other three items. Thus, an overall recall variable, memory, is created using unaided brand recall,
aided brand recall and message recall (Table 2). This variable can take either the values 0, 1,
Item Factor loading Cronbach’s Alpha
Unaided brand recall .79
.74
Aided brand recall .88
Message recall .77
Table 2. Reliability and Factor Analyses for Recall Items
Independent variable measures
Involvement is measured using one of the measures from Zaichowsky’s (1994) Personal
Involvement Inventory scale. Involvement is measured by the statement ‘This ad is relevant to me’, conforming to the definition of involvement by Zaichowsky (1985) and Wright (1974), which stress the importance of personal relevance in defining and measuring the involvement construct. Participants answered to this statement on a five-point Likert scale, where 1 = Totally Disagree, 2 = Somewhat Disagree, 3 = Neither agree nor disagree, 4 = Somewhat Agree and 5 = Totally Agree. A higher value refers to a higher level of perceived involvement, and vice versa.
Emotional Tone is measured using several statements about the perceived positivity of
the advertisements. The statements include ‘This ad gives me a positive feeling’, ‘This ad gives me a positive impression of the brand’ and ‘This ad is likeable’. Participants answered to these statements on the same five-point Likert scale as the involvement construct. Reliability analysis and factor analysis are performed to confirm that these items measure the same construct (Table 3). A factor analysis is appropriate to use, as shown by KMO measure of sampling adequacy (.75) and Bartlett’s test of sphericity (p < .001) The three items are averaged into one variable measuring Emotional Tone, with higher values indicating a more positive emotional tone, and lower values indicating a more negative emotional tone.
Item Factor loading Cronbach’s Alpha
This ad gives me a positive feeling .90
.89 This ad gives me a positive impression of the brand .90
This ad is likeable .91
Table 3. Reliability and Factor Analyses for Emotional Tone Items
Mediating variable
Viewing Time is the total amount of time the participant spent viewing the
Moderating variable
Device Category is a categorical variable indicating the device the participant used to
view the video advertisements. It originally contained five different options, where 1 = smartphone, 2 = featurephone, 3 = tablet, 4 = other mobile and 5 = desktop or laptop. None of the participants used a featurephone or a device classified under other mobile, reducing the number of devices to three. As this research aims to identify the influence of mobile versus non mobile devices, and tablets are often classified as mobile devices (Grewal et al. 2016; Stewart et al. 2019), two categories remain: 1 = Mobile device (smartphones and tablets) and 2 = Non-mobile device (desktop/laptop). The former category is used as the reference category in the analyses.
Control variables
4. DATA ANALYSIS
Before diving into the analysis, the data is checked for irregularities and outliers. Furthermore, descriptive statistics and correlations are presented, and some transformations are made to the variables.
4.1 Pre-analysis Examination and Descriptive Statistics
Upon investigation of the data, some irregularities were found. The variable ‘Viewing Time’ shows some unexpected values. As participants were only able to skip an advertisement after five seconds, the variable ‘Viewing Time’ should be equal to or higher than five seconds by definition. However, for 52 cases the variable showed values lower than 3 or showed no value at all. The time participants viewed the ad is recorded in a hidden text box, but for users with an older browser this may not work properly. It could thus be that these participants wrote a value in the text box, so that an incorrect viewing time was recorded. These 52 responses are therefore considered missing and left out of the analysis, reducing the data set to 9639 responses. Next to that, 425 observations showed a viewing time between 4.8 and 5 seconds. This is likely due to rounding errors, and therefore the values are changed to 5 seconds.
Descriptive statistics and correlations of the variables of interest are provided in Table 4. A high correlation between the two independent variables, involvement and emotional tone, is observed. To justify keeping the two variables separate, a confirmatory factor analysis is conducted. The results of a one-factor solution are compared to the results of a two-factor solution. The results show that the two-factor solution provides a better fit, as the Root Mean Square Error of Approximation (RMSEA, 0.021 vs 0.029) and the Bayesian Information Criterion value (BIC, 97382 vs. 97390) are both lower for the two-factor solution compared to the one-factor solution, indicating that the two-factor solution is a better fit. Thus, it is justified to keep involvement and emotional tone as separate variables.
1 2 3 4 5 6 7 8 9 10 11 1. Emotional Tone 1 2. Involvement .712*** 1 3. Device (mobile) -.069*** -.051*** 1 4. Viewing Time .121*** .101*** .182*** 1 5. Recall .253*** .196*** .067*** .324*** 1 6. Ad Recall .161*** .130*** .032** .124*** .253*** 1 7. Age -.104*** -.104*** .392*** .309*** .074*** .018 1 8. Gender (male) .034*** .018 -.118*** .013 -.013 -.005 -.014 1 9. Country: Germany .035*** -.006 .168*** -.131*** -.055*** -.036*** -.242*** .003 1 10. Country: UK .058*** -.002 .133*** .021* .094*** .077*** .099*** -.012 -.507*** 1 11. Gen. Attitude .409*** .482*** -.059*** .152*** .198*** .124*** -.068*** -.028** .029** .077*** 1 Mean 3.17 2.93 1.47 11.32 1.24 .85 45.13 1.51 0.33 0.34 .00 Std. Deviation 1.03 1.20 .50 5.32 1.17 .36 16.95 .50 .47 .47 1.00
4.2 Data Transformations
Some of the variables of interest are transformed before running the models. As adding interactions to the models can induce multicollinearity, the independent variables as well as the mediator are mean centered. Mean centering variables takes out multicollinearity for moderators. Even though multicollinearity should not be an issue due to this mean centering of the variables, the VIF scores of all models were checked. The highest VIF score that emerged was 2.26, indicating that the models do not suffer from multicollinearity.
The dependent variable, memory, is a variable of ordinal nature. An ordered logit model would therefore be a good fit. After fitting an ordered logit model, a brant test was conducted to test for the parallel lines assumption. This brant test was significant, even after collapsing some levels of the variables, meaning that the parallel lines assumption does not hold and that an ordered logit model cannot be used. Thus, an alternative method is needed. De Leeuw, Mair and Groenen (2017) suggest a way to transform ordinal variables into linear variables using Multivariate Analysis with Optimal Scaling (MVAOS). The technique quantifies categorical variables and transforms numerical variables to optimize the linear or bilinear least squares fit, thereby minimizing a loss function. Using the morals function, the variable memory is transformed into a linear variable, making it possible to use a linear regression analysis. The original values and their transformed values are presented in Figure 2.
Figure 2. Morals Transformation of Memory
5. RESULTS
In this section, the results of the moderated mediation analyses are presented. The results are presented separately for both independent variables (emotional tone and involvement). They are tested in separate models to clearly measure their separate influences, as by including several IV’s in a model, there is the possibility that they cancel out each other’s effect (Hayes 2017). Furthermore, results are presented for the memory variable, but also for ad recall, which is the form of memory that could not be summed with the other memory items.
5.1 Memory – Emotional Tone
The results of the analysis for emotional tone on memory are presented in Table 5a and 5b. Table 5a shows the path analyses of the mediation. Looking at the direct effects, emotional tone has a positive and significant direct effect on advertising memory (b = .0022, p < .001). This finding supports hypothesis 1. This direct effect becomes smaller when the mediator is included in the model (b = .0019, p < .001), but since it is still significant, viewing time partially mediates the relationship between emotional tone and memory. Furthermore, emotional tone has a positive effect on viewing time (b = .0682, p < .001), which in turn has a positive effect on memory (b = .0028, p < .001). Thus, hypothesis 3 is supported.
The other direct effect of interest in this research is the effect of the mobile device on memory. We find that memory is higher for non-mobile devices than it is for the reference category mobile devices (b = .0014, p < .001). This is in contrast to the hypothesis, which suggested the opposite. Therefore, hypothesis 5 is not supported.
expected. Thus, hypothesis 6 is not supported. A visualization of the effects is presented in Figures B1 and B2 of Appendix B.
Outcome Variable C: Memory A: Viewing Time B & C’: Memory
B SE B SE B SE
Emotional Tone .00216*** .00011 .06819*** .01446 .00188*** .00011 Device Category (non-mobile) .00139*** .00023 .12555*** .02175 .00102*** .00022 Emotional Tone x Device Category .05511** .01967
Viewing Time .00284*** .00014
Device Category x Viewing Time .00023 .00019
Age .00004*** .00001 .01709*** .00062 -.00001 .00001
Gender -.00024 .00020 .04991* .01885 -.00039* .00019 Country: Germany .00002 .00025 -.18702*** .02405 .00057* .00024 Country: UK .00170*** .00025 -.12861*** .02455 .00209*** .00024 General Attitude Towards Advertising .00130*** .00014 .17217*** .01358 .00078*** .00014 Table 5a. Path Coefficients for Emotional Tone. *** p < .001, ** p < .01, * p < .05
Test of Mediation B LLCI ULCI
Device Category = 1 (Mobile)
ACME .00019*** .00012 .00027
ADE .00188*** .00167 .00210
Total Effect .00208*** .00186 .00230
Prop. Mediated .09347*** .05639 .13000
Device Category = 2 (Non-Mobile)
ACME .00038*** .00029 .00047
ADE .00188*** .00165 .00210
Total Effect .00226*** .00202 .00248
Prop. Mediated .16773*** .13200 .20800
Test of Moderated Mediation B LLCI ULCI
ACME -.00019** -.00031 -.00006
Table 5b. Results of (Moderated) Mediation for Emotional Tone. *** p < .001, ** p < .01, * p < .05
5.2 Memory – Involvement
Outcome Variable C: Memory A: Viewing Time B & C’: Memory
B SE B SE B SE
Involvement .00154*** .00011 .04613** .01388 .00132*** .00011 Device Category (non-mobile) .00144*** .00023 .12809*** .02205 .00105*** .00022 Involvement x Device Category .05043* .01936
Viewing Time .00290*** .00014
Device Category x Viewing Time .00026 .00020
Age .00004*** .00001 .01705*** .00062 -.00001* .00001
Gender -.00013 .00020 .05479** .01966 -.00030 .00019 Country: Germany .00038 .00026 -.16858*** .02369 .00089*** .00024 Country: UK .00211*** .00025 -.10830*** .02506 .00245*** .00024 General Attitude Towards Advertising .00180*** .00014 .19306*** .01327 .00121*** .00013 Table 6a. Path Coefficients for Involvement. *** p < .001, ** p < .01, * p < .05
Test of Mediation B LLCI ULCI
Device Category = 1 (Mobile)
ACME .00013*** .00005 .00021
ADE .00132*** .00110 .00154
Total Effect .00146*** .00123 .00167
Prop. Mediated .09170*** .03580 .14330
Device Category = 2 (Non-Mobile)
ACME .00031*** .00021 .00040
ADE .00132*** .00111 .00153
Total Effect .00163*** .00140 .00019
Prop. Mediated .18750*** .13244 .24200
Test of Moderated Mediation B LLCI ULCI
ACME -.00017** -.00029 -.00005
Table 6b. Results of (Moderated) Mediation for Involvement. *** p < .001, ** p < .01, * p < .05
The results for the moderated mediation analysis with involvement as the independent variable are presented in Table 6a and 6b. Table 6a shows that involvement has a positive and direct effect on memory (b = .0015, p < .001), supporting hypothesis 2. Furthermore, a higher level of involvement leads to a higher viewing time (b = .0461, p = .002), which in turn increases memory (b = .0029, p < .001). These results support the hypothesis that viewing time mediates the relation between involvement and memory. This is a partial mediation, as the direct effect of involvement on memory remains significant, but smaller (b = .0013, p < .001). Thus, hypothesis 4 is supported.
Hypothesis 5 suggests that the device used has a direct effect on memory, with the effect being stronger for mobile devices than for non-mobile devices. The analysis supports the opposite, namely that memory is stronger when participants use a non-mobile instead of a mobile device (b = .0014, p < .001). Thus, hypothesis 5 is again not supported.
Table 6a also shows that the effect of involvement on viewing time is stronger for non-mobile devices than it is for non-mobile devices (b = .0504, p = .014; Figure 5), but the interaction was not significant for the relation between viewing time and memory. Even though the overall interaction is not significant, Figure 6 shows that for a higher viewing time, it does become significant, where memory becomes significantly higher for non-mobile than for mobile devices. Looking at the mediation results in Table 6b, we see that a greater proportion of the effect of involvement on memory is mediated by viewing time for non-mobile (18.8%) than for mobile devices (9.2%). This difference is found to be significant (b = -.0002, p = .004). We find that the moderated mediation is stronger for non-mobile devices, which is in contrast to hypothesis 7 which posits the opposite. Thus, hypothesis 7 is not supported. A visualization of these results is presented in Figures B3 and B4 in Appendix B.
5.3 Ad recall – Emotional Tone
As reliability and factor analyses showed that ad recall could not be summed together with the other recall items into one memory variable, this section will show the results of the moderated mediation analysis with ad recall as the variable of interest. As this is a binary variable, a logistic regression analysis is performed.
supporting hypothesis 1. This probability decreases to 41.09% in the presence of viewing time, indicating that there is a partial mediation. This can also be seen by the significant effects of emotional tone on viewing time (b = .0682, p < .001), and viewing time on ad recall (b = .3153,
p < .001). Thus, hypothesis 3 is also supported for ad recall. This mediation relationship is
stronger for non-mobile devices than it is for mobile devices. 5.9% of the effect of emotional tone on ad recall is mediated through viewing time for mobile devices, while this is 10.5% for non-mobile devices. This difference is, however, not significant (b = -.0010, p = .238). Thus, hypothesis 6 is not supported for ad recall.
Similar to the results for memory, we find that ad recall is higher for viewers using a non-mobile rather than a mobile device (b = .3087, p < .001). Specifically, the probability of correctly recalling the ad over not recalling is 36.16% higher for non-mobile device users than it is for mobile device users. As is the case for memory, hypothesis 5 is not supported.
Outcome Variable C: Ad Recall A: Viewing Time B & C’: Ad Recall
B SE Odds
Ratio
B SE B SE Odds
Ratio Emotional Tone .36609*** .03198 1.4421 .06819*** .01418 .34420*** .03224 1.4109 Device Category (non-mobile) .30868*** .06666 1.3616 .12555*** .02155 .27783*** .06846 1.3203
Emotional Tone x Device Category .05511** .01895
Viewing Time .31532*** .04528 1.3707
Device Category x Viewing Time .01053 .06399 1.0106
Age .00155 .00191 1.0016 .01709*** .00621 -.00372. .00199 .9963
Gender -.01942 .05799 .9808 .04991* .01904 -.03687 .05835 .9638
Country: Germany .04173 .07104 1.0426 -.18702*** .02423 .07133 .07133 1.0904
Country: UK .49019*** .07692 1.6326 -.12861*** .02428 .53674*** .07754 1.7104 General Attitude Towards
Advertising
.21486*** .04177 1.2397 .17217*** .01377 .16484*** .04236 1.1792
Table 7a. Path Coefficients for Emotional Tone. *** p < .001, ** p < .01, * p < .05
Test of Mediation B LLCI ULCI
Device Category = 1 (Mobile)
ACME .00256*** .00136 .00395
ADE .04091*** .03432 .04770
Total Effect .04347*** .03690 .05050
Prop. Mediated .05892*** .03146 .09000
Device Category = 2 (Non-Mobile)
ACME .00379*** .00243 .00545
ADE .03237*** .02738 .03790
Total Effect .03616*** .03076 .04170
Prop. Mediated .10487*** .06782 .14460
Test of Moderated Mediation B LLCI ULCI
ACME -.00102 -.00290 .00063
5.4 Ad Recall – Involvement
Table 8a and 8b show the results for involvement on advertising recall. The results show a similar pattern as for memory. Involvement has a positive and significant effect on ad recall (b = .2957, p < .001), and this effect becomes smaller when the mediator viewing time is incorporated in the logistic regression analysis (b = .2778, p < .001). Specifically, without the presence of viewing time, increasing involvement by one unit increases the probability of correctly recognizing the ad over not recognizing by 34.40%. In turn, when viewing time is included, this probability decreases to 32.02%. The mediation effect is significant, as involvement positively influences viewing time (b = .0461, p = .001), which in turn positively influences ad recall (b = .3206, p < .001). These results together show support for a partial mediation. Thus, hypothesis 2 and 4 are supported.
Outcome Variable C: Ad Recall A: Viewing Time B & C’: Ad Recall
B SE Odds Ratio
B SE B SE Odds Ratio
Involvement .29565*** .03179 1.3440 .04613** .01356 .27781*** .03195 1.3202
Device Category (non-mobile) .31659*** .06647 1.3724 .12809*** .02189 .28633*** .06827 1.3315
Involvement x Device Category .05043* .01928
Viewing Time .32064*** .04516 1.3780
Device Category x Viewing Time .01851 .06382 1.0187
Age .00110 .00191 1.0011 .01705*** .00062 -.00438* .00198 .9956
Gender -.00251 .05779 .9975 .05479** .01927 -.02093 .05817 .9793
Country: Germany .10497 .07102 1.1107 -.16858*** .02360 .14830* .07131 1.1599
Country: UK .56053*** .07700 1.7516 -.10830*** .02504 .60362*** .07764 1.8287
General Attitude Towards Advertising
.28647*** .04031 1.3317 .19306*** .01345 .23003*** .04097 1.2586
Table 8a. Path Coefficients for Involvement. *** p < .001, ** p < .01, * p < .05
Test of Mediation B LLCI ULCI
Device Category = 1 (Mobile)
ACME .00183** .00064 .00309
ADE .03438*** .02728 .04140
Total Effect .03621*** .02874 .04360
Prop. Mediated .05061** .01881 .08700
Device Category = 2 (Non-Mobile)
ACME .00319*** .00202 .00478
ADE .02705*** .02160 .03270
Total Effect .03024*** .02457 .03600
Prop. Mediated .10565*** .06884 .15820
Test of Moderated Mediation B LLCI ULCI
ACME -.00122 -.00284 .00047
Table 8b shows the difference in the mediation effect for both levels of the mediator. We again find that the mediation effect is stronger for non-mobile devices. 10.6% of the effect is mediated through viewing time for non-mobile devices, while this is only 5.1% for mobile devices. This difference is however not significant (b = -.0012, p = .154), thus the moderated mediation hypothesis is not supported for ad recall.
Looking at the model for the direct effects, we find that device category has a significant effect on ad recall. Specifically, the effect is higher for non-mobile than it is for mobile devices. The probability of correctly recognizing the ad over not recognizing increases by 37.24% when using a non-mobile rather than a mobile device. Thus, hypothesis 5 is not supported.
5.5 Model Fit
The validity of the models will be assessed by looking at the overall significance of the models, as well as the proportion of the variance in the variables of interest that is explained by the predictor variables. First, all models are found to be overall significant (p < .001). This means that the models for memory, both for emotional tone and involvement, explain more of the variation in memory or than would a random model.
Table 9 provides the adjusted R2 for the memory models, and pseudo-R2 measures for the ad recall models. For the memory models, the predictor variables explain 8.99% and 7.45% of the variance in memory for emotional tone and involvement, respectively. This increases to 16.20% and 15.01% when the mediator is included. For the ad recall models, three pseudo-R2 measures are analyzed. These are Cox & Snell, Nagelkerke and McFadden R2 and compare the (log)likelihood values of the model to those of a model without explanatory variables. Table 9 shows that these values are between .03 and .06 for the direct effect models. These are quite low and thus the model fit is relatively low. However, the model fit does improve when adding the mediator to the models, as then the pseudo R2 measures are between .04 and .08.
Table 9. Model Fit.
Memory Ad recall
Adjusted R2 Adjusted R2 Cox & Snell Nagelkerke McFadden
The adjusted R2 values are found to be higher for the emotional tone models than the involvement models. Thus, emotional tone is a stronger predictor of memory than is involvement. However, the R2 values are relatively low. This indicates that there are other factors that are important in explaining advertising memory. This is also highlighted by the low pseudo R2 values.
5.6 Sensitivity analysis
A strong assumption of mediation models is the sequential ignorability assumption. This is an assumption of mediation models that implies that there should be no other unmeasured variables that could affect the mediator and the outcome variables in mediation models. Assessing whether this assumption is violated is therefore necessary to assess the validity of mediation models. Sensitivity analysis allows to check the robustness of the findings with regard to the potential violation of this assumption (Imai, Keele and Yamamoto 2010).
Figure 7. Sensitivity Analysis for Emotional Tone Figure 8. Sensitivity Analysis for Involvement
emotional tone but does have a significant impact on memory. Thus, it can be concluded that the findings are relatively robust.
5.7 Predictive Validity
Figure 9. Predictive Validity for Emotional Tone Figure 10. Predictive Validity for Involvement
5.8 Control Variables
The mediation analyses also included some control variables. For age and gender, the results are inconclusive. In some of the path models, these variables seem to have a significant effect, while in others, they show insignificant results. Furthermore, the direction of the effects differs. The direction is positive for some path models, but negative for others. Therefore, it can be concluded that age and gender do not significantly influence memory of online video advertisements.
General attitude towards advertising is another control variable included in this research. In the models for emotional tone and involvement, for memory but also for ad recall, the results show positive and significant results. General attitude towards advertising thus increases memory and ad recall, but also viewing time. The effect of attitude on memory and ad recall becomes smaller but is still positive and significant when viewing time is included in the models, Thus, viewing time also seems to act as a partial mediator between the general attitude towards advertising and memory.
The last control variable is the country of origin of the participants. These are the Netherlands, Germany and the United Kingdom. Results of the analyses show inconclusive results for Germany, as the results for this country are significant in the presence of the mediator models but insignificant when the mediator is not included. However, for the UK, the results are significant. In all models, memory or ad recall is higher for participants from the UK compared to participants from the Netherlands (reference category), but viewing time seems to be lower. As these results are quite interesting, they will be further analyzed.
To investigate whether the found relationships differ per country, a separate analysis is run where dummy variables are included for the three countries, thereby leaving out the intercept. The results of the path analyses for emotional tone and involvement are presented in Table 10. The intercepts for the models with memory as the dependent variable are always lowest for participants from the Netherlands but are highest for the viewing time models. Thus, memory is lowest for participants from the Netherlands, but viewing time is the highest for this group of participants. For participants from the United Kingdom, intercepts for memory are the highest, but for German participants, intercepts for viewing time are the lowest. Thus, results for these two countries remain inconclusive.
for the models for emotional tone and involvement. However, only for Germany a significant moderated mediation is found. The results of the path analyses and the mediation analyses per country are presented in Appendix C.
Outcome Variable C: Memory A: Viewing Time B & C’: Memory
B SE B SE B SE Emotional Tone Netherlands -.00289*** .00035 -.74692*** .03403 -.00069. .00035 Germany -.00286*** .00030 -.93395*** .02817 -.00011 .00031 United Kingdom -.00118** .00035 -.87554*** .03281 .00140*** .00035 Involvement Netherlands -.00315*** .00035 -.76258*** .03389 -.00087* .00035 Germany -.00278*** .00030 -.93116*** .02873 .00003 .00032 United Kingdom -.00105** .00035 -.87087*** .03376 .00158*** .00034
Table 10. Intercepts for the Path Analyses per Country
Test of Mediation Netherlands Germany UK
Emotional Tone ACME .00026*** .00019** .00036*** ADE .00186*** .00171*** .00211*** Total Effect .00213*** .00190*** .00247*** Prop. Mediated .12412*** .10100** .14427*** Involvement ACME .00016*** .00018** .00028*** ADE .00134*** .00131*** .00131*** Total Effect .00150*** .00149*** .00159*** Prop. Mediated .10700*** .12023** .17305***
Test of Moderated Mediation
Emotional Tone: ACME -.00012 -.00034** -.00014 Involvement: ACME -.00008 -.00039** -.00008
6. DISCUSSION
As online video advertising is an increasingly popular form of digital advertising, more research is needed to examine the drivers of the effectiveness of these ads. This paper added to the advertising literature in a few key ways. This paper has investigated the relationship between two creative elements, emotional tone and involvement, in relation with an important measure of advertising effectiveness, namely memory of advertisements and advertised brands. It took into account viewing time as a mediator and investigated if this effect differed for different devices used to view the advertisement. A summary of the results is presented in Table 12. These findings have some important theoretical and practical implications, which will be discussed next.
Hypothesis Hypothesized
direction
Supported? Direction found
H1: Emotional Tone → Memory + ✓ +
H2: Involvement → Memory + ✓ +
H3: Emotional Tone → Viewing
Time → Memory
+ ✓ +
H4: Involvement → Viewing Time
→ Memory
+ ✓ +
H5: Device Category → Memory Memory higher for mobile devices
Memory higher for non-mobile devices
H6: Emotional Tone: Moderated
Mediation
Mediation effect stronger for mobile devices
Mediation effect stronger for non-mobile devices Not significant for ad recall
H7: Involvement: Moderated
Mediation
Mediation effect stronger for mobile devices
Mediation effect stronger for non-mobile devices Not significant for ad recall
Table 12. Summary of the Results of the Hypotheses.
6.1 Theoretical Implications
This paper makes important contributions to the advertising literature, and specifically, to the online video advertising literature. It has already been demonstrated before that creative elements such as emotional tone and involvement have a positive effect on advertising effectiveness and memory in particular. This paper has added to the literature by demonstrating that this relationship also holds for skippable, online video advertisements. Specifically, the relationship is found to be somewhat stronger for the emotional tone than for involvement.
video advertisements on platforms such as YouTube are usually skippable after only a few seconds. It is thus important to know what influences this viewing time.
Third, this paper sheds light on the influence of the device used to view an online video advertisement. Little prior research has examined this relationship. While it was hypothesized that users viewing the advertisement on a mobile device would have a better memory of the ads and advertised brands, the opposite was found. There are a few explanations as to why memory is higher for non-mobile devices such as laptops and desktops. One of the few papers examining the influence of the device, Stewart et al. (2019), suggests that digital video advertising is more effective for laptop users than it is for mobile device users as laptop screen sizes are larger. Though they do not provide a theoretically grounded argument for this, it could be that when the screen is larger, the advertisement is also larger and thus captures more of the viewer’s attention. Thus, memory is higher for non-mobile devices as the larger video size captures more of the viewer’s attention.
Another contribution is that this paper takes into account the device as a moderator of the mediation relationship. Here, it was also hypothesized that the mediation relationship would be stronger for mobile devices than it would be for non-mobile devices. Although a significant moderated mediation effect was found for advertising memory, for both emotional tone and involvement, the effect was again in the opposite direction as hypothesized. Again, an explanation for this can be that screen size is the most important factor in determining a device’s advertising effectiveness.
6.2 Practical Implications
Adding to this, marketers should also make use of the finding that viewing time partially mediates the relationship between emotional tone and involvement on memory. As video advertisements are usually skippable after only a few seconds, it is crucial to capture a viewer’s attention in these first few seconds. This paper demonstrated that when a video ad has a positive emotional tone and a higher perceived involvement, the viewing time increases. This highlights the importance of incorporating these elements in online video advertisements.
Another important implication is that memory is higher for participants viewing the advertisement on a non-mobile device such as a laptop rather than a smartphone or tablet. Marketing practitioners can use this finding when determining how to allocate advertising budgets. As advertising effectiveness is found to be higher for non-mobile devices than for mobile devices, allocating a higher share of the marketing budget towards non-mobile devices increases effectiveness of campaigns.
6.3 Limitations and Further Research
A few limitations of this research will be discussed next. The first limitation is related to the indicator used to measure involvement. To measure perceived involvement, only one indicator is used. Even though this indicator relates directly to the definition of involvement, it is only measured using one item and therefore has a lower reliability. Previous research has measured involvement using the ten-item construct known as the Personal Involvement Inventory developed by Zaichowsky (1994). This is a measure of involvement known for its high validity and reliability. Due to unavailability of data, this could not be used in this research. Further research should therefore consider using the Personal Involvement Inventory to measure perceived involvement.
Another reason to use the Personal Involvement Inventory is that in this research, involvement and emotional tone are highly correlated. Even though it was proven that they are separate constructs, they do however overlap to some extent. This overlap will likely be reduced when using a more reliable construct of involvement.
question, this thought is not unlikely. Further research should therefore consider a different study design where the questions are more strictly separated, to avoid this Halo Effect.
The effect sizes that were found were rather small, both for the direct effects and the mediation effects. One explanation for this is that the transformation of the memory variable led to very small values, but this does not explain all. It is therefore likely that there are other factors at play that influence participants’ memory of online video advertisements. This is highlighted by the relatively low R2 values that were found. An example would be to investigate whether the video advertisements are perceived as humorous (Chung and Zhao 2003) and whether this influences memory. Also, it can be examined whether an interaction between the two independent variables, emotional tone and involvement, exists. Furthermore, this research found that participants’ general attitude towards advertising has a positive impact on advertising memory. It is possible that part of this effect is mediated through viewing time. Thus, further research should examine the influence of other variables on advertising memory.
7. REFERENCES
Allan, David (2007), “Comparative Effectiveness of 30- versus 60-Second Radio Commercials on Recall and Rate,” Journal of Radio Studies, 14 (2), 165-77.
Belanche, Daniel, Carlos Flavián, and Alfredo Pérez-Rueda (2017), “Understanding interactive online advertising: Congruence and product involvement in highly and lowly arousing, skippable video ads,” Journal of Interactive Marketing, 37, 75-88.
Bolls, Paul D., Annie Lang, and Robert F. Potter (2001), “The effects of message valence and listener arousal on attention, memory, and facial muscular responses to radio advertisements,” Communication Research, 28 (5), 627-51.
Brasel, Adam S., and James Gips (2014), “Tablets, Touchscreens, and Touchpads: How Varying Touch Interfaces Trigger Psychological Ownership and Endowment,” Journal of
Consumer Psychology, 24 (2), 226-33.
Buchholz, Laura M., and Robert E. Smith (1991), “The Role of Consumer Involvement in Determining Cognitive Response to Broadcast Advertising,” Journal of Advertising, 20 (1), 4-17.
Cacioppo, John T., Wendi L. Gardner, and Gary G. Berntson (1997), “Beyond bipolar conceptualizations and measures: The case of attitudes and evaluative space,” Personality and Social Psychology Review, 1 (1), 3-25.
Chung, Hwiman, and Xinshu Zhao (2003), “Humour effect on memory and attitude: moderating role of product involvement,” International Journal of Advertising, 22 (1), 117-44. Eckler, Petya, and Paul Bolls (2011), “Spreading the virus: Emotional tone of viral advertising and its effect on forwarding intentions and attitudes,” Journal of Interactive Advertising, 11 (2), 1-11.
Fortin, David R., and Ruby Roy Dholakia (2005), “Interactivity and vividness effects on social presence and involvement with a web-based advertisement,” Journal of Business Research, 58 (3), 387-96.
Grewal, Dhruv, Yakov Bart, Martin Spann, and Peter Pal Zubcsek (2016), “Mobile Advertising: A Framework and Research Agenda,” Journal of Interactive Marketing, 34, 3-14.
Hayes, Andrew F. (2017), Introduction to mediation, moderation, and conditional process analysis: A regression-based approach, 2nd ed. New York: The Guilford Press.
IAB (2020a), “IAB Video Advertising Spend Report” (accessed October 23, 2020), https://www.iab.com/wp-content/uploads/2019/04/IAB-Video-Advertising-Spend-Report-Final-2019.pdf
IAB (2020b) “Internet Advertising Revenue Report” (accessed October 15, 2020),
https://www.iab.com/wp-content/uploads/2020/05/FY19-IAB-Internet-Ad-Revenue-Report_Final.pdf
Imai, Kosuke, Luke Keele, and Teppei Yamamoto (2010), “Identification, inference and sensitivity analysis for causal mediation effects,” Statistical Science, 25 (1), 51-71.
Joa, Claire Youngnyo, Kisun Kim, and Louisa Ha (2018), “What Makes People Watch Online In-Stream Video Advertisements?” Journal of Interactive Advertising, 18 (1), 1-14.
Keller, Kevin Lane (1993), “Conceptualizing, Measuring, and Managing Customer-Based Brand Equity,” Journal of Marketing, 57 (1), 1-22.
Lang, Peter J. (1979), “A bio-information theory of emotional imagery,” Psychophysiology, 16 (6), 495-512.
Lang, Annie (2000), “The Limited Capacity Model of Mediated Message Processing,” Journal
of Communication, 50 (1), 46-70.
Lang, Annie (2009), “The Limited Capacity Model of Motivated Mediated Message Processing,” in The Sage Handbook of Media Processes and Effects, R.L. Nabi and M.B. Oliver, eds. Thousand Oaks, CA: Sage Publications, 193-204.
Lang, Annie, and Kulijinder Dillon (1995), “The effect of emotional arousal and valence on television viewers’ cognitive capacity and memory,” Journal of Broadcasting and Electronic
Media, 39 (3), 313-27.
Leuthesser, Lance, Chiranjeev S. Kohli, and Katrin R. Harich (1995), “Brand equity: the halo effect measure,” European journal of marketing, 29 (4), 57-66.
Li, Hao, and Hui-Yi Lo (2015), “Do You Recognize Its Brand? The Effectiveness of Online In-Stream Video Advertisements,” Journal of Advertising, 44 (3), 208-18.
Moorman, Marjolein, Peter C. Neijens, and Edith G. Smit (2007), “The effects of program involvement on commercial exposure and recall in a naturalistic setting,” Journal of
Advertising, 36 (1), 121-37.
Olney, Thomas J., Morris B. Holbrook, and Rajeev Batra (1991), “Consumer responses to advertising: The effects of ad content, emotions, and attitude toward the ad on viewing time,” Journal of Consumer Research, 17 (4), 440-53.
Petty, Richard E., and John T. Cacioppo (1981), Attitudes and Persuasion: Classic and
Contemporary Approaches, Dubuque, IA: Wm. C. Brown.
Petty, Richard E., John T. Cacioppo, and David Schumann (1983), “Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement,” Journal of
Consumer Research, 10 (2), 135-46.
Reeves, Byron, R., John Newhagen, Edward Maibach, Michael Basil, and Kathleen Kurz (1991), “Negative and positive television messages: effects of message type and message context on attention and memory,” American Behavioral Scientist, 34 (6), 679-94.
Peltier, James W., and John A. Schibrowsky (1994), “Need for cognition, advertisement viewing time and memory for advertising stimuli,” ACR North American Advances, 21, 244-50.
Song, Sigen, Fanny Fong Yee Chan, and Yanlin Wu (2019), “The interaction effect of placement characteristics and emotional experiences on consumers’ brand recognition,” Asian
Pacific Journal of Marketing, 32 (6), 1269-85.
Stewart, Kirsten, Matt Kammer-Kerwick, Allison Auchter, Hyeseung Elizabeth Koh, Mary Elizabeth Dunn, and Isabella Cunningham (2019), “Examining digital video (DVA) effectiveness,” European Journal of Marketing, 53 (11) 2451-79.
Teixeira, Thales, Michel Wedel, and Rik Pieters (2012), “Emotion-induced engagement in internet video advertisements,” Journal of Marketing Research, 49 (2), 144-59.
Varan, Duane, Jamie Murphy, Charles F. Hofacker, Jennifer A. Robinson, Robert F. Potter, and Steven Bellman (2013), “What works best when combining television sets, PCs, tablets, or mobile phones?: how synergies across devices result from cross-device effects and cross-format synergies,” Journal of Advertising Research, 53 (2), 212-20.
Wright, Peter (1973), “Cognitive Processes Mediating Acceptance of Advertising,” Journal of
Marketing Research, 10 (1), 53-62.
Zaichowsky, Judith Lynne (1985), “Measuring the Involvement Construct,” Journal of
Consumer Research, 12 (3), 341-52.
APPENDIX A: SURVEY
Figure A1. Example of Video Advertisement as shown to Participants
Figure A2. Survey Question for Unaided Recall
Figure A4. Survey Question for Message Recall
Figure A5. Survey Question for Ad Recall
