The effect of humor in online live-chat

The effect of humor in online live-chat

Master Thesis

MSc. Business Administration – Track Marketing University of Amsterdam, Faculty of Economics and Business

Author: Kelly Hagebeek Student number: 11094206 Supervisor: Hsin-Hsuan Meg Lee

Statement of originality

This document is written by Student K.R. Hagebeek who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

Live-chat is becoming more important and popular, as it is an effective tool to interact with consumers online. Despite the importance of interaction online, little is known about the concept of live-chat. Interaction increases the consumer engagement and brand attitude, but there are several ways in communication to address consumers in such an interaction. Humor for example can have contrary effects, but is often expected to have a positive effect. This study investigates the willingness to engage in a conversation and the brand attitude when humor is used to address consumers in a proactive live-chat. In addition, it is investigated whether brand personality and the online setting influence the effect of humor on the consumer behaviors. The results of an online experiment, involving 242 respondents, showed no support for a positive effect of humor on the consumer engagement and brand attitude. Furthermore, no evidence was found for the suggested moderating effects of the brand personality and online setting. Yet, this research presents encouraging support for the suggestion that both moderators may help explain the differences in consumers behavior. As, the consumer engagement and brand personality are more positive with a warm brand personality than with a competent brand personality. Also, a public setting elicited a more positive brand attitude when humor was used than in a private setting, whereas respondents were less willing to engage in a public setting than in a private. Finally, this study provides theoretical and managerial implications and concludes by giving suggestions for future research.

Contents

1. Introduction ... 5

1.1 Thesis overview ... 9

2. Literature review ... 10

2.1.1 Interaction and consumer engagement ... 12

2.1.2 Interaction and brand attitude ... 13

2.2 Live-chat strategies ... 14

2.3 Humor in live-chat and consumer responses ... 15

2.4 Influence of brand traits on the perception of humor ... 17

2.5 Online setting in live-chat and consumer responses ... 19

3 Research Design ... 23 3.1 Procedure ... 23 3.2 Manipulation ... 24 3.3 Measure ... 26 3.3.1 Dependent variables ... 26 3.3.2 Control variables ... 27 3.3.2 Manipulation check ... 29 3.4 Pre-test ... 30 3.5 Sample ... 31

3.6 Data analysis strategy ... 33

3 Results ... 34 4.1 Data preparation ... 34 4.2 Descriptive statistics ... 34 4.3 Manipulation check ... 35 4.4 Correlation matrix ... 37 4.5 Hypotheses testing ... 39 4.5.1 Moderating effects ... 41 5 Discussion ... 47

5.1 Discussion of the results ... 47

5.2 Theoretical implications ... 51

5.3 Managerial implications ... 52

5.4 Limitations and future research ... 53

6. Conclusion ... 55

7. References ... 57

1.! Introduction

Nowadays, consumers have created higher expectations from brands in the electronic commerce (e-commerce) business according to Kang et al. (2014). Consumers want more than just browsing through the webpages online. Apparently, brands need to add more value online to satisfy these expectations. But what does more value consists of? Interaction can be an important factor to create more value. Online interaction has a great influence on a consumers’ response to a website (Agarwal and Venkatesh, 2002; Jiang and Benbasat, 2007). According to some studies, interactivity even has a positive influence on consumers’ responses (McMillan et al., 2003) and on the consumers’ general attitude of the website (Wu ,1999). It is also established that with the interaction with consumers online, brands can gain positive customer relationships and build up brand product loyalty (Guo et al., 2010). Even more, it is shown that online purchase intentions have reduced by the absence of the online interaction with a brand or product representative (Jiang et al., 2010; Qiu and Benbasat, 2005). A way of implementing this interaction is through the use of a live-chat medium on websites and social media (Kang et al., 2014).

The live-chat medium is becoming more popular as more companies are recognizing the importance of this feature and implementing it on their website or using it on social media (Elmorshidy, 2013; Jiang et al., 2010; Kang et al., 2014). A live-chat is an online medium which companies can use to socially interact with consumers by means of a chat. This medium makes it possible to communicate and engage with the consumer, for example to answer questions, make suggestions or help to finalize the sale by chatting directly with the consumer (Roggio, 2009). In this way, the consumer receives an instant response from the

2013). Furthermore, a live-chat medium is a cost-effective tool to bring customization and social interaction to the expected online customer experience and increases the responsiveness of consumer questions. As provided in the literature, interaction has a positive influence on brand attitude and consumer’s willingness to engage with the brand. Based on this, we believe that brand attitude and consumer engagement are important factors to implement in this study.

A live-chat medium is often implemented on shopping websites with the goal to inform, interact, persuade and offer personalized service for consumers (Kang et al., 2014; Macias, 2003). Examples are large online retailers such as Amazon and Taobao (Kang et al., 2014). However, a live-chat can also be implemented on social media platforms of brands with the goal to respond, engage, interact and connect directly with consumers (Heller Baird and Parasnis, 2011). Examples are brands that have a social media account, such as Facebook, and are using this as a live-chat function where they communicate with consumers directly, like KLM and L’Oréal Paris.

There are certain elements concerning live-chat that can be taken into consideration, like the way of addressing consumers (the communication style) and the online environment or setting in which the interaction of such a live-chat takes place. It is known that privacy is of great importance for a positive relationship between the brand and the consumer (Bart et al., 2005). Also, consumers care about their privacy online. Some consumer are afraid of losing control of their privacy, hence they do not engage in public sharing for example or are not willing to share everything publicly (Han and Maclaurin, 2002). On the other hand, Stephen et al., (2015) stated that people are more inclined to interact with others online when questions are being asked for example, here no distinction is being made whether this is publicly or privately.

Two kinds of virtual settings can be distinguished for a live-chat, which are important for this study. The first one occurs in a ‘private online setting’, where the interaction is merely between the brand (representative) and the consumer with them being the only one to see the interaction. The second one is set again between the brand (representative) and the consumer, but other consumers or even brands can still see and participate in the interaction if they want to. This setting is referred to as a ‘public online setting’.

Effective communication will add value to the interactions (Ingram, 1992). Packard et al., (2014) stated that the matter of impact certain words have on interactions between a consumer and a brand representative has been ignored in the literature. Although, a few researches (Packard et al., 2014; Lenoir, 2015) looked at effective communication and how companies should address consumers in an interaction. Examples of this are the use of personal pronouns and formality, which both proved to be influencing consumer responses. The use of the ‘I’ pronoun in the interaction with the consumer has a positive influence on consumer’s attitude of the brand (Packard et al., 2014). For consumer responses of the use of formality, it is even shown that formality and brand personality are connected. It is revealed that consumers preferred an informal address when the brand was associated with the brand personality warmth, whereas consumers preferred and elicited more positive responses when a formal way of addressing was associated with a competent brand (Lenoir, 2015). So the personality of a brand can influence the consumer’s attitude to the interaction.

Another communication style used to address consumers is humor. Humor satisfies the need of social connection (Martin, 2010) and it increases a positive brand attitude (Gulas and Weinberger, 2006). Much research is conducted for the use and the benefits of humor, though humor that is set-up through online communication has not been explored yet (Shifman and Blondheim, 2010). Many large brands are using humor as it believes to be

we focus on the relationship between humor and the brand attitude and the willingness to engage in a conversation.

Taking the elements of live-chat into consideration, we are going to examine how humor plays a role in creating consumer engagement and a positive attitude towards the brand via an online live chat function and whether these relationships are moderated by the brand’s personality and the online setting. In this way, this research makes a theoretical contribution. First of all, it fills the research gap of the concept of live-chat and the effect of humor on brand attitude and consumer engagement, while looking at two brand traits and the virtual setting that moderates this relationship in live-chat. We shed light on the behavior of consumers when humor is used in the address to consumers in live-chat. We found out whether they are willing to engage in a conversation and whether they have a positive brand attitude after experiencing this form of addressing in live-chat. Additionally, we provide insights in how the brand personality and online setting influences this engagement and attitude. And it also complements recent studies by expanding the research towards the live-chat medium.

Lastly, this study is also of practical relevance and value to firms and managers. We can provide guidance for managers who are willing to implement such a live-chat function on their website. More specifically, we can help them with creating the most effective chat-up line, in terms of the usage of humor. If firms and managers know how and when to use humor in the chat-up line they will have the opportunity to increase the attitude towards the brand and the willingness to engage in conversation with the brand positively. In addition, it is important for them to learn what kind of influence humor has on the brand attitude and consumer engagement, while consumers are on their website or social media platform.

The research question is: ‘What effect does the use of humor in live-chat have on brand attitude and consumer engagement and how do brand traits and the virtual setting moderates this relationship?’

1.1 Thesis overview

This research is structured as follows. In the next chapter a review of the literature on the key concepts about live-chat, humor, consumer engagement, brand attitude, live-chat strategies, brand personality (warmth and competence) and the virtual setting are being discussed. Furthermore, a theoretical framework about the relationship between humor and brand attitude and the willingness to engage in a conversation is developed, as well as for the moderating variables virtual setting and brand personality. The hypotheses are explained and formulated and the conceptual framework is presented. In the third chapter the research design and methodology is explained. The results of the research will then be presented in chapter five. Hereafter, the empirical findings are discussed, the managerial implications are explained and the limitations and suggestions for further research are given. Finally, a conclusion is provided.

2.! Literature review

This chapter provides a comprehensive review of the literature about interaction, humor, brand personality and the virtual setting in this study, in order to analyze what already has been studied about the factors that have an effect on the brand attitude and the willingness to engage by means of an online live chat. First of all, general information and previous research about live-chat and interaction in the online environment is given, followed by theory and information about the relation between live-chat and consumer engagement and live-chat and brand attitude. Hereafter, the independent factor humor is being analyzed and previous research about the virtual setting and brand traits relating to the address of consumers is being reviewed. Lastly, based on this review the research gap and the research question are identified, the hypotheses and a conceptual framework are developed and tested in this study.

2.1! Live-chat

A live-chat medium is an online communication tool that provides the opportunity to let the consumer and a seller or brand representative interact real-time about product and service information (Ou and Davison, 2009). From the brand’s point of view, this medium is an additional service for the consumer in comparison with the traditional customer service as it creates a direct or instant personal relationship between the consumer and the brand representative online (Kang et al., 2014). According to Goes et al., (2011) the availability of this communication tool provides the opportunity to replicate the experience of physical store shopping, where real life communication or interaction is possible.

The live-chat function is becoming more popular because it is being recognized by companies (Andrews, 2010). Live-chat may even become a trend in the business world as it is believed that live-chat can increase customer satisfaction. It can be used directly to assist consumers and it can respond to the need of social interaction online by consumers.

Making use of a live-chat medium can increase sales. For example, in the study of Dukcevich (2002) it is illustrated that the company named Lands’ End has a chat function on the website called Live Help, which has been associated with their e-commerce success. When a customer uses the Live Help tool, the average value of an order increases by 6%. Their sales rose as they found out that 20% of their new customers were coming from the internet and that the costumer who made use of the live-chat function are 20% more likely to purchase online then customer who did not used the live-chat function. We can link this increase in sales to the increase in use of the live-chat function, which could have generated a higher customer satisfaction or a positive attitude towards the website and the brand.

Shae et al., (2007) conducted a survey of customer satisfaction and the use of a live-chat medium. According to this study, one of the benefits of a live-live-chat function are customer care efficiency and cost reduction. However there are also some challenges presented like ‘scheduling and routing functionality, archival of problem resolution sessions, unification with knowledge management systems, and efficient interfaces for agents to effectively handle and multi-task several chat sessions’ Shae et al. (2007, p. 1).

Nevertheless, Ou et al. (2008) concluded that live-chat improves the consumer’s view of interactivity and interaction has proven to be effective for maintaining and building a customer relationship (Qiu and Benbasat, 2005; Kalyanam and Zweben, 2005; Aberg and Shahmehri, 2000) and . This can also be found in the study of Andrews (2010), which stated that most companies that have this live-chat function on their website, use this tool ‘as a

cost-to consumers questions and personalization of the shopping experience’ Andrews (2010, p. 1).

We can state that the purpose of live-chat is to interact with consumers. There are two specific elements which are of great important to companies when they set up an interaction by means of a live-chat; consumer engagement and brand attitude.

2.1.1! Interaction-and-consumer-engagement-!

Interactivity is also an important factor for the consumer engagement. By posting content online, companies hope to interact with consumers and to trigger their interest which cause more consumer engagement with the brand (Stephen et al., 2015). According to several studies (e.g Brodie et al., 2013; Laroche et al., 2013), it is advantageous to gain a high consumer engagement as it is believed to be related with increased purchasing, loyalty, satisfaction and brand affinity and a stronger consumer-brand relationship. What does consumer engagement even mean in context of a live-chat? And would an interaction online by means of a live-chat increase the consumer engagement?

Consumer engagement has many different definitions in the literature. Especially in the online context. Here, consumer engagement can include posting, subscribing, favoriting, distributing, networking and emailing. Because the level of user interaction is one of the most important factors when consumer engagement is being measured (Ghuneim, 2008), we define, for the purpose of this study, consumer engagement as the willingness to engage into a conversation with the brand. This can be elaborated in the context of live-chat online.

As being discussed in the paper of Stephen et al., (2015), a positive effect has been found between comments of consumers and the feedback questioning of brands on Facebook; as a reaction to questions or comments by the company. As they explain it: ‘it is normal to

ask questions and for people to answer those questions’ Stephen et al. (2015, p.44). So it seems that there is a positive relationship between the branded content online and the consumers’ engagement with it. However, a social media post of a brand could only receive more comments of consumers (so more engagement) when the particular post meet certain social communication standards. For instance, asking the consumer’s about their opinion and not telling the consumer to actually do something because this could elicit a negative effect (Stephen et al., 2015). For this reason, we state that the way of addressing consumers online, either on social media or on a brand’s website is an important factor.

2.1.2-Interaction-and-brand-attitude-Another reason for companies to implement such a live-chat medium on their website or social media platform is to increase the consumers’ website or social media platform satisfaction. This satisfaction will consequently lead to a more positive attitude towards the brand itself. So, consumers who engage in online interaction with messages and consumers who interact online with other humans in a way of sender –receiver interaction, will evaluate the website more positive, which in turn will lead to a positive attitude towards the brand (Ko et al., 2005). From this is can be concluded that online interaction, with either humans or just messages, can elicit a positive attitude towards the brand.

The attitude towards the brand is most of the time influenced by marketing campaigns and can be even more influenced by direct marketing. Moreover, ‘’The attitude toward the brand encompasses the extent to which the firm is able to create close connections or emotional ties with the consumer’’ Lemon et al., (2001, p.22). With a live-chat, such a connection or tie can be made because of the interaction function (Ou et al., 2008).

Considering the fact that live-chat increases the brand attitude positively, this study is going to examine in what way the live-chat can be used to create a positive brand attitude.

2.2-Live6chat-strategies-There are two types of live-chat strategies that companies can use depending on who initiated the interaction first. The first one is using a reactive chat system which means that the consumer has to take the initiative of chatting with a brand representative by clicking on the chat button on the website or social media platform. Examples of this reactive live-chat system on a website can be found on the websites of the Weddingshop and The Little Green Bag: http://www.weddingshop.com/, https://www.thelittlegreenbag.nl/. The second strategy is based on a proactive live-chat, which means that the brand representative takes the initiative to chat with the consumer, by inviting them to chat or interact with them by means of a pop-up chat window or an interactive inviting post on social media. Most previous studies (Shae et al., 2007; Yin and Straub, 2002) focused on a reactive chat system. But when a proactive approach was investigated in a service setting, it resulted in an optimal customer relationship management (Köhler et al, 2011). However, proactive live-chat present different issues which not have been discussed much in the literature. Potential issues could be negative consumer outcomes as a result of privacy invasion feelings and failure of response when the consumer reacts (Challagalla et al., 2007). For this reason it is worth investigating the influences of a pro-active live-chat in this study.

So when brands make use of a pro-active live-chat, they are the one who addresses the consumer. Consumers may not respond to the live-chat, but does the attempt of communicating with the consumer increase the brand attitude? And in what way will brands try to address the consumer with a live-chat to create that positive brand attitude. Many things

strategy. The way of addressing, or the language used in addressing consumers is important for the effectiveness of the marketing communication (Lenoir, 2015). Factors that could be taking into consideration could be the use of formality, pronouns, language and different emotions. For example, more and more companies are using humor on social media in an attempt to increase the relationship with the consumer (Lee et al., 2015). But does that also mean that using humor to address consumers with a live-chat medium increase the brand attitude?

All in all, we want to examine how to make use of an effective proactive communication interaction to create a positive attitude towards the brand as well as the increase in the willingness to engage into a conversation, when humor is used. For this reason we provide a literature background for the humor in the next section.

2.3!

Humor-in-live6chat-and-consumer-responses-Humor is a kind of emotion that is recommended for marketing. It provides multiple benefits like the positive attitude of consumers towards the communication involving humor, the increase in attention of consumers and it is beneficial for the brand attitudes (Gulas and Weinberger, 2006; Eisend, 2011).

As such, humor is used by brands in advertising and on social media. Warren and McGraw (2013 ) explain that humor is used in advertising because it is believed to increase the likelihood of purchase and recommendation of the brand by consumers, which results in enhancing brand attitudes. As for social media, Stephen et al., (2015) even states that humor in posts on Facebook can decrease the negative attitude of consumers towards the content of the post. Looking at these benefits of humor, we establish that humor is an important factor

Next to this, humor is often used in conversations as a function for ‘breaking the ice’ and to reduce a barrier for communication (Goldstein, 1982). Based on this, we believe that humor has a positive influence on the willingness to engage into a conversation. Also, as to our knowledge so far, humor has not been investigated in relation to use of humor in live-chat. We hypothesize the following:

H1: Humor used in live-chat will lead to a positive consumer engagement. H2: Humor used in live-chat will lead to positive brand attitudes.

In the literature, humor is divided into four different styles. Affiliative and self-enhancing are two positive types, whereas aggressive and self-defeating are two negative types (Dyck and Holtzman, 2013). This research focusses on the positive affiliative humor style because of the following explanation. Affiliative humor is often used with the goal to amuse others, to reduce interpersonal stress and perhaps most importantly to maintain or even strengthen the relationship in a positive way (Martin et al., 2003). Or as DeLongis & Holtzman (2005) explain it, this humor is relationship-focused and is associated with satisfaction and a high level of support. An example of this style is telling a joke or playful teasing. This will elicit positive reactions which supports a more intimate and harmonious relationship (Dyck and Holtzman, 2013). As such, we focus on this style of humor because it could influence the relationship between the consumer and the brand positively when the brand addresses the consumer.

Besides the positive effects of humor, research of Warren and McGraw (2013) has shown that humor used in marketing sometimes can actually be harmful to brand attitudes. When the attempt or use of humor fails to the extent that the consumer does not find it humorous, this will elicit negative feelings which will damage the brand. This can also happen with positive humor because of potential misinterpretation risks or the limited

possibility to transfer the humor intention, which is in live-chat only possible with text (Hancock, 2004). Therefore we can conclude that humor used in marketing communication is an essential focus point for managers. With this research we want to provide managers with the outcome of using humor in online marketing communication in relation to the attitude towards their brand and the willingness to engage in an online conversation with the brand.

2.4! Influence of brand traits on the perception of humor

There are some factors that could influence the way consumers evaluate humor when considering engagement and brand attitude. Examples of these factors could be the use of formality, consumer personality, brand personality and even the setting online.

‘’Brand personality is understood as the human characteristics or traits that can be attributed to a brand’’ Keller and Richey (2006, p. 74). When a brand communicates with a consumer, the consumers tries to recognize the brand’s personality (Lenoir, 2015) as consumers transfer human personality traits onto brands. This transfer, along with the advertising strategies of personalizing the brands by companies, can be beneficial for brands as it can differentiate them from other companies (Aaker, 1997).

It is argued by Lenoir (2015), that the characteristics of a brand (or the brand personality) influences the consumers’ attitude, reactions to address and even the willingness to buy. She specifically argues that brand personality affects the consumer preference to the use of formality. In this study, humor is used to address consumers by means of a live-chat, which could be in conflict with a competent brand personality. Furthermore, Aaker et al., (2010) found out that the consumers’ perceptions of companies are strongly influenced by the assessment of warmth and competence. Based on this, we suggest that brand personality moderates the relationship between humor and consumer engagement and brand attitude and

H3: Brand personality will be a moderator of the relationship between the use of humor and the consumer engagement and brand attitude.

According to Fiske et al., (2007) there are two universal principles of social perceptions, namely warmth and competence. Humans perceive brands based on these two perceptions (Kervyn et al. ,2012). Friendliness, approachability and trustworthiness are the related traits of warmth, whereas ability, skill and efficacy represents the related traits of competence (Lenoir, 2015). Many research in the social psychology, organizational behavior and branding presented findings that consumers distinguish others on the traits of warmth and competence (Aaker et al., 2010; Fiske et al., 2007; Judd et al., 2005). Prior research (Lenoir, 2015) suggest that warmth and competence moderated the impact of the preferred way of addressing address. However, in a live-chat condition this influence is not yet known. For this reason our research will include these two brand traits, warmth and competence.

According to Martin et al. (2003), the tendency for people to engage in an interaction with affiliative humor reflects personality traits like warm and kind. This indicates that when the personality trait warm is perceived, affiliative humor will be perceived positively. Meanwhile, warmth indicates the intentions towards the consumer which is based on emotions, making humor more appropriate (Kervyn et al. ,2012). Based on this, we expect that a ‘warm’ brand personality will have a more positive influence on the relationship humor and consumer engagement.

On the other hand, humor may not always be appropriate and could be perceived differently by the consumer in a different setting and even with different brand personalities. For example, brands that are perceived to have a competent personality reflect skill, ability and efficiency. The focus here is on whether the company is capable enough to carry out his intentions and not what the intentions are towards the consumer, which is more based on emotions (Kervyn et al. ,2012), therefore humor will not be appropriate and may even have a

negative effect. Lenoir, (2015) also showed more positive responses from consumers when a formal way of addressing was used with a competent brand than a warm brand. So because humor is viewed as unprofessional in formal situations (Dziegielewski, 2003), we argue that a competent brand personality will lead to less positive consumers behaviors when humor is used.

Concerning the two brand personalities, warmth and competence, present in this study and their expected influence on consumer behaviors when humor is used, we present the following hypotheses:

H3a: The brand trait warmth will lead to a positive relationship between humor and consumer engagement and brand attitude.

H3b: The brand trait competence will lead to a weaker relationship between humor and consumer engagement and brand attitude.

2.5! Online setting in live-chat and consumer responses

Two types of online settings can be distinguished when an online interaction occurs between the consumer and a brand. First, a ‘private’ setting is explained. Here, a consumer can have an one-on-one interaction with a representative of a brand via a live-chat which is not visible for others. So, this situation is between two people and personal. Imagine a live-chat window pops up on the brands’ website or consumers receive a message in their private message inbox from the brand on Facebook. The second setting is explained as a ‘public’ setting. Here, a consumer can have an interaction with the brand via a live-chat that is publicly available and visible for others. This means consumers who are posting, commenting and responding to question/comments of brands on their Facebook page for example.

Would consumers have a different brand attitude or consumer engagement when brands use humor to address consumers on the web for the whole world to see (public setting), or on the website of the brand in a more private setting? We expect that there would indeed be a difference in consumer engagement and brand attitude. On the one hand, consumers might prefer reacting to brands privately to avoid possible public information sharing (Han and Maclaurin, 2002). On the other hand, consumers may want to react to a brand that uses humor publicly to show to everyone how funny they are, which is caused by the need to express their self-esteem (Cheung et al., 2011). According to Berger and Milkman (2012), humor is a type of amusement that is classified as a positive high-arousal emotion. They conclude that when a positive high-arousal emotion, or in this case humor, is evoked consumers are more likely to discuss and react to the content publicly. This suggests a positive influence in a public setting for the consumer behaviors when humor is used.

Recent studies on social media are focusing more on the public setting, as they imply that nowadays most interaction occur publicly (Heller Baird and Parasnis, 2011; Mangold and Faulds, 2009). Research suggested that consumers participate in self-presentational behavior, trying to increase their self-esteem when they interact publicly (Krämer and Winter, 2008; Cheung et al., 2011). Also, when consumers interact publicly they can manage their self-presentation more strategically since they are in control in if they want to respond and when they want to respond. In a private setting, this control is less as a more immediate response is expected since the interaction is now direct and more personal (Krämer and Winter, 2008). Hence, we suggest that the setting can have different influences on the consumer’s behavior and a public setting will influence the consumer behaviors positively: H4: Virtual setting will be a moderator of the relationship between the use of humor and the consumer engagement and brand attitude.

H4a: A public setting will lead to a positive relationship between humor and consumer engagement and brand attitude.

Furthermore, interactions can be quick and efficient in private settings. But when a consumer feels that the personal interaction is inappropriate, it can actually backfires (Connell et al., 2001). In these kind of situations, humor can perhaps be used to reduce the interpersonal stress (Martin et al., 2003), serve as an ‘icebreaker’ and even lower this barrier to communicate (Goldstein, 1982). However, there is one essential concept in online interaction which contradicts the fact that a humor can be used in a private setting and that is the online privacy. Privacy is one of the most important factors of trust, and consumers need trust to be able to build a favorable relationship with the brand (Bart et al., 2005). But when this private factor creates a form of trust with consumers, is humor then even appropriate to address consumers? More privacy is provided in a private setting than in a public one and most consumers care about their privacy. They do not want to share personal information publicly and they are afraid of losing control of their privacy (Han and Maclaurin, 2002). Moreover, a private setting can also be experienced more formal and in most formal settings humor is not appropriate as it is received as unprofessional (Dziegielewski, 2003). As no answer to this question was found in the live-chat literature, we provide the following hypotheses:

H4b: A private setting will lead to a weaker relationship between humor and consumer engagement and brand attitude.

Now that we have reviewed the majority of the literature about live-chat and the relating factors, the objective is to relate this to a business context. As mentioned before there is a rise in live-chat systems, however despite of the rise of this live-chat feature, there is limited research on this online communication tool (Kang et.al, 2014). Taking this and the literature above into consideration, we think it is relevant to propose the following research question which is related to addressing consumers in a live-chat; ‘What effect does the use of humor in live-chat have on brand attitude and consumer engagement and how does brand traits and the virtual setting moderates this relationship?’. The conceptual model of this research can be found in figure 1.

3! Research Design

To investigate what effect the use of humor has on the willingness to engage in a conversation and brand attitude in brand chat-up lines and how brand traits and the virtual environment moderates this relationship, research is conducted and data is collected. The overall design of this study will have an experimental setting and the research will be conducted by the use of an online questionnaire. More specifically, this research consists of a Vignette study. Vignette studies offer insights of how the individual’s thoughts, behavior, feelings and decisions are influenced by certain factors, which can be manipulated by the researcher by displaying different scenarios in the questionnaire (Evans et al., 2015).

A Vignette study fits the purpose of this study because of the hypothetical scenarios given to the respondents online. A 2X2X2 factorial design was displayed to elicit thoughts, general impressions, willingness etc. regarding the brand which will be proposed in the different scenarios. Qualtrics was used to provide the online questionnaire.

Before the survey was distributed, a pre-test was conducted to establish whether the manipulations of humor were actually perceived funny or humorous and whether the brand personality was manipulated in the right manner.

The brand which is used for this study is a fictitious online food delivery service, that goes by the name Food Now. The reason for the use of this fictitious brand is to prevent prejudgments of the respondents which may bias them. Furthermore, Facebook was chosen to be implemented in the survey as it is the most used social media platform online (Rank, 2016), hence probably the most easiest to identify for consumers.

3.2 Manipulation

The respondents chosen must have had access to internet and have visited or have the intention to visit the website or social media platform of a brand.

The 2X2X2 factorial design consist of two settings for humor, humorous and non-humorous, two settings for brand personality, warm and competent and two settings for the online context, public setting and private setting (see appendix 1 for all the scenarios and corresponding questions of the survey).

The respondents are presented with the scenarios first, then they were asked to what extent they find the brand competent or warm and to what degree they found it humorous. Furthermore, they were asked about their willingness to engage and their attitude towards the brand. Finally, they had to answer whether they could imagine this scenario happening in real life before they received general questions about their online behavior and demographical questions.

The first scenario, where the relation of humor on the willingness to engage and the brand attitude, is tested in a public virtual environment in terms of a chat on social media with the brand trait warmth. The second scenario is presented in a private virtual environment in terms of a private online chat, where the relation of humor on the willingness to engage and the brand attitude is tested with the brand trait warmth. The third scenario where the same relation is tested, is presented in a public virtual environment with the brand trait competence. The fourth scenario includes a private virtual environment and the brand trait competence, whereas again the relation of humor on the willingness to engage and the brand attitude is tested. Lastly, these four scenarios are also conducted to the relation between non-humor and the willingness to engage and the brand attitude. Which makes a total of eight different scenarios. The scenarios with the corresponding questions are presented in the appendix.

Three main manipulations are used to conduct this study. Humor, brand personality and the online setting were manipulated in the scenarios. Humor was manipulated by the use of an affiliative humor style. The following line was used to indicate the humor used by the brand: 'Our mission is to keep you from having to cook, shop or wear pants! Order today, leftovers tomorrow. Did you just solve making dinner twice?'. This line involves affiliative humor because it is joke telling or playful teasing, as it is a funny thing to say, to amuse the consumer and it is order to provide extra value (next to not have to shop or cook, also not necessary to wear pants) to the consumer in a funny way (Dyck and Holtzman, 2013). Also, an exclamation mark was added in the sentence to stress the chance of perceived humor in this line (Hancock, 2004). The scenarios without humor were provided with the following line: ‘Our mission is to keep you from having to cook or to shop. Can I help you find your dinner for tonight?’.

Concerning brand personality, two kinds of brand personality were manipulated namely warmth and competence. Both were manipulated by the means of text and with the design of the brand logo provided in the scenarios. The text explained certain characteristics, or synonyms of characteristics, of that brand personality which was based on the articles of Lenoir (2015) and Aaker et al. (2010). For warmth, the friendliness characteristic was stressed as the following; ‘Food Now is known for their friendliness. Last year they have won the award of most friendly food delivery service of Europe.’ For competence, the characteristic of reliability was stressed as the following; ‘Food Now is known for their reliability. Last year they have won the award of most reliable food delivery service of Europe.’ Next to this, the logo was designed with two different colors to manipulate the personality of the brand. According to Fraser and Banks (2004), the color pink can be linked to sincerity, warmth and nurturing, whereas the color blue can be linked to competence,

reliability and efficiency. For this reason, the brand logo of Food Now was pink in the ‘warm’ personality scenario and the logo was blue in the ‘competent’ personality scenario.

The last factor to manipulate was the virtual setting. These manipulations were done by means of a text; ‘They post the following message on their public Facebook wall’, which indicates the public setting. The private setting was indicated as the following; ‘You receive the following message in your private Facebook message inbox.’ The manipulations of these descriptions are quit explicit as they even include the words describing the scenario; ‘public’ and ‘private’. Both manipulations focused on the way the respondents received or viewed the message online.

3.3 Measure

The different variables are explained hereafter. First, a literature research was conducted with the purpose of finding existing appropriate measurement instruments. The relevant measures were used when they were available. However, some adjustments need to be made in wording to fit the context of this study.

3.3.1! Dependent variables

In this study, two different dependent variables are measured: consumer engagement and brand attitude. First, respondents viewed the manipulated content. This includes the brand logo, which was either manipulated as warm or competent depending on the scenario and a description of the brand which was also manipulated either as a warm or competent brand. Then the consumer engagement was measured on a four-item, seven-point Likert scale (1=strongly disagree and 7=strongly agree) (Hollebeek et al., 2014), whereby the respondents responded on the following statements: ‘I am willing to comment on this message’, ‘I am willing to answer the question’, ‘I am willing to engage with this brand online’, ‘I am willing

to engage into an online conversation with this brand’. The items were averaged to form a total consumer engagement score on this scale and proven to be reliable; α = .918.

Next, the brand attitude was measured again after respondents viewed the manipulated logo and description of the brand. This measurement was done on a four-item, seven-point differential scale (1= negative, bad, unpleasant, unfavorable and 7=positive, good, pleasant, favorable) (Puzakova et al., 2013), whereby respondents answered the question what their attitude or feeling towards the brand is after experiencing this scenario. These items were averaged to form a total brand attitude score on a reliable scale of α =.972.

3.3.2 Control variables

Realistic check

A realistic check question was added in this study to check whether the respondents perceive the scenarios as realistic. When respondents do not perceive the scenario as realistic, chances are that they will fill in randomly chosen answers in the survey instead of answering according to what they actually believe or act like. For this reason, the name and the logo of the fictitious brand were carefully picked to represent in this study, as existing online food delivery services were compared. In the survey, there was only one item that was measured; ‘I can imagine this scenario happening in real life’, on a seven-point Likert scale (1=strongly disagree and 7= strongly agree). Because there was only one item to measure, a realistic check was not possible to conduct.

Online behavior

The respondents were presented with three questions about the online behavior. The questions asked about the respondents’ time spent online, how often they visit a company’s website or Facebook online and how often they interact with a brand when they are online. Two examples were given for interaction; ‘commenting on their Facebook post or by chatting with one of the employees on their website’. For the first question respondents could choose between ‘0-30 minutes’, ’30-60 minutes’, ‘1-2 hours’, ‘2-3 hours’, ‘>3 hours’, ‘none at all’. For the second question, the respondents had to choose between ‘Always’, ‘Most of the time’, ‘About half the time’, ‘Sometimes’, ‘Never’. For the final question the possibilities were again ‘Always’, ‘Most of the time’, ‘About half the time’, ‘Sometimes’, ‘Never’.

Demographics

The final questions that were asked to the respondents were about the demographics. These questions provide insights of the respondents. This can be of aid to determine what kind of factors can be of influence on the respondents’ thoughts and answers. First it was asked to indicate the age by the following categories; ‘Under 18’, ’18-24’, ’25-34’, ’35-44’, ’45-54’, ’55-64’ and ’65 or older’. Next, the gender was determined by the options ‘Male’ and ‘Female’. Last, the level of education was asked according to one of the following; ‘High School’, ‘Intermediate vocational education (MBO)’, ‘Bachelor (HBO)’, ‘Bachelor WO’, ‘Master’.

3.3.2! Manipulation check

Although three manipulations exist in this study, only two manipulations checks were included. The first and most important one is humor, followed by the brand personality. The third manipulation which is not included in the manipulation check is the online setting. The reason for this is that the online setting was manipulated by means of a description in the scenario, explaining the manipulation. Respondents were directly told whether they were situated in the public or private scenario which was quit straightforward, instead of priming indirectly. Hence, no manipulation check was needed.

To find out whether the respondents find the proposed humor actually humorous/funny, respondents were asked to what degree they find the manipulation

humorous, funny and to what degree it puts a smile on the face. The independent variable was measured on a three-item, seven-point Likert scale (1=not at all and 7=very much) (Hansen et al., 2009). All these items were averaged to form a total humor score on this scale and proved to be reliable; α = .924 and N=130. The other 112 respondents received a scenario for non-humor, which was measured on the same scale which again was reliable, α = .914. The total score of humor and non-humor were tested together to check whether the total of humor and non-humor is reliable. Results showed that they are reliable, α = .921.

The purpose of the second manipulation check is to verify whether respondents actually perceived the warm manipulated scenarios as warm and the competent manipulated scenarios as competent. This item was measured with a six item, seven-point Likert scale (1=not at all and 7=very much). The following traits indicated the warm brand personality: warm, kind, generous. Whereas, the traits indicated the competent brand personality were: competent, efficient, effective (Aaker et al., 2010). By showing the fictitious brand name and the manipulated logo the following questions were presented: ‘To what extent do you believe

extent do you believe that Food Now is generous?’, ‘To what extent do you believe that Food Now is competent?’, ‘To what extent do you believe that Food Now is efficient?’, ‘To what extent do you believe that Food Now is effective?’. The items were recoded and averaged to form a total warmth and total competence score of α =.919 (warm) and α =.856 (competent). This implicates that the item are reliable.

3.4 Pre-test

A pre-test was conducted among 22 respondents (54,5% was male). The pre-test was conducted to ensure that the experimental manipulation of the scenarios worked, also the pre-test was used to check the reliability of the different scales. A one sample t-pre-test was conducted to check whether the manipulated scenarios worked. The values were testes against the mean of the scale (4).

The respondents scored statistically different for the warm scenario (M=5.83, SD=0.73) than the average of the scale t(21)= 11.73, p<.001 and the respondents received a statistically different score for competent (M=5.21, SD=0.87) than the value of four t(21)=6.49, p<.001.

Next, the respondents found the realistic scenarios statistically different scenario (M=4.95, SD=0.85) than the average of the scale t(21)= 5.24, p<.001.

Finally, non-humor was received with a statistically difference (M=2.4, SD=1.08) than the value of four t(21)=-6.9, p<.001. However, humor was not found to be of significant difference (M=4.38, SD=1.52) than four t(21)= 1.18, p=.255. Adjustments had to be made for the manipulation of humor as this is the dependent variable hence the most important manipulator of this study. The line that was used to manipulate humor in the pre-test was: ‘Our mission is to keep you from having to cook, shop or wear pants! Can I help you find you

cook, shop or wear pants! Order today, leftovers tomorrow. Did you just solve making dinner twice?' To check whether this line has a significant difference for humor, another one sample t-test was conducted. The result of this t-test showed a significant difference for humor (M=4.78, SD=1.54) than four t(20)= 2.32, p=.031. Therefore, the latter line was used in the final survey to manipulate humor.

Furthermore, a reliability analyses in SPSS showed that all scales (brand personality, humor, willingness to engage, brand attitude and realistic check) were reliable as all Cronbach’s Alpha (from now on α) scores were bigger than .7.

3.5 Sample

Convenience sampling was used to acquire respondents. These respondents were gathered randomly and consists of students and acquaintances of the researcher. Altogether, 276 respondents filled in the questionnaire. However, some data was missing as 34 respondents did not fill in the questionnaire completely. Therefore, this data had to be withdrawn from the dataset. So, the total sample size of this research is 242 respondents (50% men and 50% women) of which 38% of all respondents fell into the age group of 25-34 years old and 38.8% of all respondents fell into the age group of 18-24 years. There were only 5 respondents in the age group of ‘Under 18’, 0 in the group of ’60 or older’ and only 4 respondents in the group of ’65 or older’. Because these small amounts are inappropriate for a proper analysis, the latter two groups were merged with ’55-64’ and the former were merged with the age group of ’18-24’. Also, 37.6% of all the respondents have the master degree as the highest completed (or current) level of education. Most of the respondents (31%), spends 30-60 minutes on average on social media per day, whereas 64% only visits a company’s website or Facebook page when they are online sometimes (on a scale of 1= never and

5=always). Lastly, 50.4% of all the respondents indicated that they never interact with a brand when they are online. Descriptive data of the sample can be found in table 1.

Factor N % Gender Male 121 50 Female 121 50 Age 18-24 94 38.8 25-34 92 38 35-44 28 11.6 45-54 28 11.6

Education High school 16 6.6

Intermediate vocational education (MBO)

38 15.7

Bachelor (HBO) 57 23.6

Bachelor WO 40 16.5

Master 91 37.6

Time spent on social media 0-30 minutes 64 26.4

30-60 minutes 75 31

1-2 hours 56 23.1

2-3 hours 31 12.8

>3 hours 12 5

None at all 4 1.7

How often visits on

website/Facebook of brand

Always 7 2.9

Most of the time 27 11.2

About half the time 33 13.6

Sometimes 155 64

Never 20 8.3

How often interaction with brand online

Always 2 0.8

Most of the time 3 1.2

About half the time 11 4.5

Sometimes 104 43

Never 122 50.4

3.6-Data-analysis-strategy-To be able to analyze the obtained data and to present results, several tested needed to be conducted. First, two kind of t-tests were conducted for the manipulation checks. The one sample t-test was able to verify whether the respondents scored significantly different on an item compared to the mean of the item scale. The mean of the scale was four in all of the scenarios, because the scale presented one to seven. Second, the independent sample t-test was used to compare the means between two manipulated scenarios on the same variable. For example, for brand personality it was used to compare the means between the warm and competent scenario on the variable warmth. Also, this test was used for humor to compare the means between humor and non-humor and verify is they significantly differ.

Finally, the hypotheses were tested by means of an analysis of covariance

(ANCOVA). For the first two hypotheses the ANOCOVA was conducted with humor as the independent variable and consumer engagement and brand attitude separately as the

dependent variables. The covariates that were added in every ANCOVA were age, gender and education. For the third hypothesis, this test was conducted with humor and brand personality as the independent variables and consumer engagement and brand attitude

separately as the dependent variables. For the final hypothesis, the ANCOVA was conducted with humor and the online setting as the independent variable and consumer engagement and brand attitude separately as the dependent variables.

3!

Results

4.1Datapreparation

-Before any analyses can be made with the obtained data, this data need to be set in the right way in order to execute the proper analysis. To do so, variables had to be computed into new variables that contain the mean value of all the same items together. Also, the control variable Age was adjusted. Only 5 respondents were in the age group of ‘Under 18’ which is too small for an analyses. For this reason we merged the ‘Under 18’ group and the ’18-24’ group. The same applies for ’65 or older’ group. This group was merged with ’55-64’.

4.2-Descriptive-statistics-!

To present the key variables used in this study, an overview is provided in the table 2 below. Here, specific results for each key variable can be found.

N Mean Median Std. Deviation

Variance Minimum Maximum Competent 118 4.81 5.00 1.33 1.77 1.00 7.00 Warmth 124 5.01 5.33 1.37 1.87 1.00 7.00 Humor 242 3.73 3.67 1.58 2.51 1.00 7.00 Consumer engagement 242 3.36 3.25 1.48 2.19 1.00 7.00 Brand attitude 242 4.43 4.75 1.42 2.01 1.00 7.00 Table 2: Descriptive statistics key variables

4.3! Manipulation check

Brand personality

Manipulation checks are conducted to make sure that the manipulations worked. A one sample t-test was conducted to determine if a significant difference existed between the brand personalities and the mean of the scale, which was the value of four. The people who viewed the warm scenario received statistically different scores for warm (M=5.02, SD=1.37) than the average of the mean t(123)=8.27, p<.001. Furthermore, the people who viewed the competent scenario received statistically different scores for competent (M=4.81, SD=1.33) than the average of the mean t(117)=6.66, p<.001. For both brand personalities it can be concluded that the manipulations worked. The respondents indeed interpreted the competent brand for competent and the warm brand for warm.

Moreover, an independent sample t-test showed for warm a significant difference between the warm manipulated scenarios (M=5.02, SD=1.37) and the competent manipulated scenarios (M=4.20, SD=1.40; t(240)=-4.59, p<.001). The results for competent showed a significant difference between the competent manipulated scenarios (M=4.81, SD=1.33) and the warm manipulated scenarios (M=4.48, SD=1.10; t(240)=2.14, p=.033). These results showed that the scores for warm in the warm manipulated scenarios were higher than in the competent manipulated scenarios, and that the scores for competent were higher in the competent manipulated scenarios than in the warm manipulated scenarios. This proves that the manipulation worked.

What is remarkable though, is that the score values are all above the test value of four. Even when the measure is not manipulated. To test whether the measures which are not manipulated are significantly different from four, a one sample t-test is conducted. The

significant different (M=4.20, SD=1.40) than the value of four t(117)=1.53, p=.129. The scores of competent in warm manipulated scenarios did show a significant difference (M=4.48, SD=1.10) than the value of four t(123)=4.87, p<.001. Hence, this result suggest that the warm manipulated scenarios worked.

Humor

A one sample t-test was conducted to check whether there was a significant difference between the humorous and non-humorous scenarios and the mean of the scale, which was the value of four. The scores for humor in the humorous manipulated scenario were not significantly different (M=4.07, SD=1.55) than the average of the mean t(129)=0.546, p=.586. Whereas the people who viewed non-humor received a significant difference for non-humor (M=3.32, SD=1.53) than the average of the mean t(111)=-4.70, p<.001. An explanation for this might be that it is difficult to find a kind of humor that everybody perceives as humor. For this reason, an independent samples t-test was conducted to check whether the humor variable significantly differed from the non-humor variable. This test did show a significance difference between humor (M=4.07, SD=1.55)and non-humor (M=3.32, SD=1.52; t(240)=-3.79, p<.001) and that non-humor is significantly higher than non-humor. Hence, the humor manipulated scenario was perceived with more humor than the non-humor manipulation, which proves that the manipulation for humor works.

Realistic

For the final manipulation, it was tested whether the respondents viewed the scenarios described as realistic by means of a one sample t-test. The results showed a significant difference between the scores of realistic (M=4.82, SD=1.49) and the value four, which is the average of the mean t(241)=8.58, p<.001). Consequently, the manipulation of reality works.

4.4!Correlation matrix

The correlations that were measured are displayed in table 3. There are several significant correlations that need to be discussed. First, a significantly positive correlation can be seen between the two dependent variables, consumer engagement and brand attitude (r=.598, p<.01). This suggest that a positive brand attitude increases the consumer engagement and that a higher consumer engagement increases the brand attitude.

Another notifiable significant correlation is between humor and the dependent variables. A positive correlation can be found between humor and consumer engagement (r=.530, p<.01) and humor and brand attitude (r=.630, p<.01). This correlation is consistent with H1a and H1b which suggested that humor has a positive effect on consumer engagement and brand engagement.

Lastly, humor positively correlates with warmth (r=.197, p<.01)., as well as with competent (r=.258, p<.01). This correlation suggest that both warmth and competence increase humor, next to the effect of humor increases warmth and competence.

! ! Age$ Gender$ Education$ Warmth$ Competent$ Realistic$ Humor$ Consumer$$ engagement$ Age$ Correlation!Coefficient! 1! ! ! ! ! ! ! ! $ Sig.!(25tailed)! .! ! ! ! ! ! ! ! Gender$ Correlation!Coefficient! 50,11! 1! ! ! ! ! ! ! $ Sig.!(25tailed)! 0,087! .! ! ! ! ! ! ! Education$ Correlation!Coefficient! 5,266**! ,210**! 1! ! ! ! ! ! $ Sig.!(25tailed)! <.001! 0,001! .! ! ! ! ! ! Warmth$ Correlation!Coefficient! 50,112! 0,016! ,168**! 1! ! ! ! ! $ Sig.!(25tailed)! 0,081! 0,806! 0,009! .! ! ! ! ! Competent$ Correlation!Coefficient! 5,131*! 0,046! ,234**! ,320**! 1! ! ! ! $ Sig.!(25tailed)! 0,042! 0,475! <.001! <.001! .! ! ! ! Realistic$ Correlation!Coefficient! 50,024! 0,017! 0,025! ,242**! ,352**! 1! ! ! $ Sig.!(25tailed)! 0,712! 0,796! 0,696! <.001! <.001! .! ! ! Humor$ Correlation!Coefficient! 0,012! 0,047! 0,073! ,197**! ,258**! ,354**! 1! ! $ Sig.!(25tailed)! 0,852! 0,47! 0,26! 0,002! <.001! <.001! .! ! Consumer$engagement$ Correlation!Coefficient! ,131*! 50,069! 5,155*! ,272**! ,237**! ,390**! ,530**! 1! $ Sig.!(25tailed)! 0,042! 0,287! 0,016! <.001! <.001! <.001! <.001! .! Brand$attitude$ Correlation!Coefficient! 50,001! 50,022! 0,051! ,365**! ,362**! ,447**! ,630**! ,598**! ! Sig.!(25tailed)! 0,988! 0,732! 0,43! <.001! <.001! <.001! <.001! <.001!

Table 3: Correlation matrix

4.5!Hypotheses testing Consumer engagement

This research proposes that humor has a positive influence on the willingness to engage (H1). To test this hypothesis, an analysis of covariance (ANCOVA) was conducted with humor as the independent variable and consumer engagement as the dependent variable. The covariates present gender, age and education. First, an one-way ANOVA was

conducted to show that there is a significant difference in consumer engagement between non-humor (control group) and humor F(1, 240)=.108, p=.044. The mean scores of the dependent variable can be found in table 4.

Next, the ANCOVA was performed including the covariates. This test presented that the consumer engagement in the non-humor group (M=3.56) was higher than in the humorous group (M=3.18), but not significantly different F(1, 237)=3.581, p=.060. Hence, H1 was not supported. The ANCOVA results for the consumer engagement including the covariates are presented in table 5. This table also shows that none of the covariates is significant.

Humor (N=130) Non-humor (N=112)

Mean SE Mean SE Consumer engagement 3.18 1.55 3.56 1.37 Brand attitude 4.46 1.44 4.39 1.39

Source Type ш Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 26.706ª 4 6.677 3.162 .015 .051 Intercept 82.341 1 82.341 39.000 .000 .141 Age 6.870 1 6.870 3.254 .073 .014 Gender .271 1 .271 .128 .720 .001 Education 4.369 1 4.369 2.070 .152 .009 Humor 7.562 1 7.562 3.581 .060 .015 Error 500.384 237 2.111 Total 3256.688 242 Corrected Total 527.090 241 Brand attitude

The second hypothesis proposed that humor also has a positive influence on the brand attitude (H2). Another ANCOVA test was performed with humor as the independent variable, brand attitude as the dependent variable and the same covariates as described above. First, by the means of a one-way ANOVA test, it is shown that there is a

significant difference in the brand attitude scores between the humor and the non-humor group F(1, 240)=1.33, p=715. Hereafter, the covariates were added and an ANCOVA test indicated that the scores for brand attitude suggest to be higher for the humor group (M=4.46) than for the non-humor group (M=4.39). However, these scores are not significantly different when the means were adjusted for the covariates F(1, 237)=0.095,

Table 5: ANCOVA results for H1.

a. R Squared = .051 (Adjusted R Squared = .035) Dependent variable: total score on willingness to engage

p=758. Thus, this finding does not provides support for H2. Table 6 provides the results for the ANCOVA. No significance has been found for the covariates in this table.

Source Type ш Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 3.800ª 4 .950 .469 .758 .008 Intercept 114.275 1 114.275 56.463 .000 .192 Age .914 1 .914 .452 .502 .002 Gender .089 1 .089 .044 .834 .000 Education 3.354 1 3.354 1.657 .199 .007 Humor .192 1 .192 .095 .758 .000 Error 479.658 237 2.024 Total 5229.938 242 Corrected Total 483.458 241 4.5.1 Moderating effects Brand personality

This research presents that brand personality has a moderating effect on the relationship between humor and the consumer engagement and brand attitude (H3). H3 was again tested by means of an ANCOVA, with humor and brand personality as the independent variables, consumer engagement and brand attitude as dependent variables and gender, age and education as covariates. The scores of the means can be found in table 7.

Table 6: ANCOVA results for H2.

a. R Squared = .008 (Adjusted R Squared = -.009) Dependent variable: total score on brand attitude

Humor (N=130) Non-humor (N=112) Warm Competent (N=64) (N=66) Warm Competent (N=60) (N=52) Consumer engagement 3.39 2.89 (1.56) (1.51) 3.40 3.76 (1.45) (1.72) Brand attitude 4.53 4.39 (1.39) (1.50) 4.15 4.67 (1.57) (1.12)

First, the main effect of the moderator (brand personality) was tested with the dependent variable consumer engagement. It was indicated that there was a significance difference F(1,138)=4.150, p=.043. Next, the ANCOVA (including the covariates) showed that consumer engagement differ for the brand personalities in the non-humor group (M= =3.40 versus M=3.76) and the humorous group (M=3.39 versus M=2.89). However, the results show an insignificant difference F(1, 235)=3.551, p=.061, which is in contrast with H3. Hence, H3 is rejected. The results of this test including consumer engagement as the dependent variable, brand personality as moderator and the covariates, can be found in table 8. Source Type ш Sum of Squares df Mean Square F Sig. Partial Eta Squared Corrected Model 34.363ª 6 5.727 2.732 .014 .065 Intercept 81.745 1 81.745 38.987 .000 .142 Age 6.391 1 6.391 3.048 .082 .013 Gender .163 1 .163 .078 .780 .000 Table 8: ANCOVA results for H3 consumer engagement

Dependent variable: total score on consumer engagement

Table 7: Dependent variable scores by humor and brand personality.