1 T HE R OLE OF A FFECT AND C OGNITION IN THE I MPACT OF POSITIVE/NEGATIVE ONLINE CONSUMER REVIEWS ON B RAND ATTITUDE AND PURCHASE INTENTION By JINSOO KIM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 2
2 201 2 Jinsoo Kim
3 To Dr. Y. Kim, a passionate scholar, respected educator, and beloved father (1940 2009)
4 ACKNOWLEDG MENTS I thank my parents and my wife, Eun young for their love, support and patience I would also like to give many thank s to my advisor, Dr. Jon D. Morris for his invaluab le guidance, mentorship, and s upport Special thanks are also in order to the mem bers of my committee for giving me invaluable advice, and insightful comments : Dr. Robyn Goodman Dr. Cynthia Morton and Dr. Yong Jae Ko I would also like to take this opportunity to thank all of the faculty members at the University of Florida, especial ly Dr. John Sutherland for his mentorship and care. I could not have completed this study without them. Thank you!
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 ABSTRACT ................................ ................................ ................................ ..................... 9 CHAPTER 1 INTRODUC TION ................................ ................................ ................................ .... 11 The Rise of WOM in the Internet Era ................................ ................................ ...... 11 Need for the Study ................................ ................................ ................................ .. 13 2 LITERATURE REVIEW AND HYPOTHESES ................................ ........................ 16 Classic Studies in WOM ................................ ................................ ......................... 16 eWOM ................................ ................................ ................................ ..................... 18 Distinct Characteristics of eWOM versus WOM ................................ ...................... 19 eWOM and Attitude ................................ ................................ ................................ 20 The Concept of Attitude ................................ ................................ .......................... 22 Tripartite Attitudinal Structure ................................ ................................ ................. 23 Cognition ................................ ................................ ................................ .......... 24 Affect ................................ ................................ ................................ ................ 24 Conation ................................ ................................ ................................ ........... 26 Attitude Formation: Cognition and Affect ................................ .......................... 27 Interplay between Cognition and A ffect ................................ ............................ 28 Affect and Cognition: Predicting Behavioral Intention ................................ ....... 30 Attitude Formation under Various Conditions ................................ ................... 32 3 M ETHODS ................................ ................................ ................................ .............. 39 Overall Procedure ................................ ................................ ................................ ... 39 Pretest 1 Manipulation 1 (Product Type) ................................ ....................... 41 Pretest 2 Manipulation 2 (Involvement) ................................ ......................... 43 Pretest 3 Manipulation 3 (Valence of Online Consumer Review) .................. 44 Main Experiment (Main Study I & II) ................................ ................................ 46 Measurement Instruments ................................ ................................ ...................... 47 Affective Respon se toward Consumer Reviews ................................ ............... 47 Cognitive Response toward Consumer Reviews ................................ .............. 47 Attitude toward Consumer Reviews ................................ ................................ 49 Consumer Review based Attitude toward Product/Brand ................................ 49 Behavioral (Purchase) Intention ................................ ................................ ....... 49
6 Analysis Strategy ................................ ................................ ................................ .... 50 M ultiple R egression A nalys i s ................................ ................................ ............ 50 SEM (Structural Equation Modeling) ................................ ................................ 50 4 R ESULTS ................................ ................................ ................................ ............... 54 Main Study I ................................ ................................ ................................ ............ 54 Manipulation Checks ................................ ................................ ........................ 54 Hypothesis Testing ................................ ................................ ........................... 55 RQ1, RQ4. Consumer Review Attitude ( A r ) Formation (H5 ~ H8) .................... 56 Hypothesis 5 (Func tional product + High involvement) .............................. 56 Hypothesis 6 (Functional product + Low involvement) ............................... 57 Hypothesis 7 (Hedonic product + Hig h involvement) ................................ 57 Hypothesis 8 (Hedonic product + Low involvement) ................................ .. 57 RQ2, RQ4. Product Attitude ( A p ) Formation (H9 ~ H12) ................................ .. 58 Hypothesis 9 (Functional product + High involvement) .............................. 58 Hypothesis 10 (Functional product + Low involvement) ............................. 58 Hypothesis 11 (Hedonic product + High involvement) ............................... 59 Hypothesis 12 (Hedonic product + Low involvement) ................................ 59 RQ3, RQ4. Purchase Intention ( PI ) Formation (H13 ~ H16) ............................ 60 Hypothesis 13 (Functional product + High involvement) ............................ 60 Hypothesis 14 (Functional product + Low involvement) ............................. 60 Hypothesis 15 (Hedonic product + High involvement) ............................... 60 Hypot hesis 16 (Hedonic product + Low involvement) ................................ 61 Main Study II ................................ ................................ ................................ ........... 62 Overall Relationship among Constructs ................................ ........................... 62 Reliability and validity ................................ ................................ ................. 63 Confirmatory Factor Analysis (CFA) ................................ .......................... 64 Structural Equation Modeling (SEM): H1, H2, H3, H4 ................................ ...... 65 5 S U MMARY AND D ISCUSSIONS ................................ ................................ ........... 76 Summary ................................ ................................ ................................ ................ 76 Think Positive, Feel Negative! ................................ ................................ ................ 79 6 LIMITATIONS AND SUGGESTIONS ................................ ................................ ...... 84 APPENDIX ................................ ................................ ................................ .................... 86 LIST OF REFERENCES ................................ ................................ ............................... 90 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 100
7 LIST OF TABLES Table page 3 1 Paired t test of product type ................................ ................................ ............... 52 3 2 Salient attributes of grammar and game software ................................ .............. 52 3 3 t test of involveme nt level ................................ ................................ ................... 52 3 4 t test of valence of online consumer reviews ................................ ...................... 52 3 5 Combinations of conditions for 8 cells ................................ ................................ 53 4 1 Consumer Review Attitude ( Ar ) f ormation ................................ .......................... 68 4 2 Product Attitude ( A p ) formation ................................ ................................ .......... 69 4 3 Purchase Intention ( PI ) formation ................................ ................................ ....... 69 4 4 Descriptive statistics of measurement items ................................ ....................... 70 4 5 Correlation matrix of measur ement items ................................ ........................... 70 4 6 Summary statistics and correlation among constructs ................................ ........ 71 4 7 Results of CFA (Group number 1 Default model) ................................ ............. 72 4 8 Fit Indices of competing models ................................ ................................ ......... 74 4 9 Standardized path coefficients of competing models ................................ .......... 75
8 LIST OF FIGURES Figure page 2 1 The Two step Flow Model ................................ ................................ .................. 36 2 2 The Hierar chy of Information Sources ................................ ................................ 37 2 3 AdSAM (Self Assessment Manikin) ................................ ................................ .. 37 2 4 Hypothesized model a nd hypotheses ................................ ................................ 38 4 1 Final measurement model ................................ ................................ .................. 73 4 3 Final model with standardized path coefficients ................................ ................. 75
9 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy T HE R OLE OF A FFECT AND C OGNITION IN THE I MPACT OF POSITIVE/ NEGATIVE ONLINE CONSUMER REVIEWS ON B RAND ATTITUDE AND PURCHASE INTENTION By Jinsoo Kim May 2012 Chair: Jon D. Morris Major: Journalism and Communications Along with the prosperity of the Internet, WOM (or eWOM) has become one of the most powerful fo rces emerging in marketing today as it is widely accepted by consumers as a critical information source. However, there has been little formal research published in scholarly journals until recent years especially to understand the process of how consumer s form their attitudes towards a brand/product based on online consumer reviews. The goal of the current study is two fold. The study investigate s how affective and cognitive processing two of the main components in forming an attitude play roles in t he impact of online consumer reviews on brand/product attitude formation, and consequently on purchase intention formation The study develop s and tests a conceptual model to explain this process by using the Structural Equation Modeling (SEM) technique. The study also examine s responses formed by their thinking about the product/brand under various conditions (e.g. product type, level of involvement, and valence of eWOM messages).
10 The study found the predominant dir ect influence of affe ct on consumer review attitude compared to cognition. It also detected the role of affective response as a mediator for cognitive response. The findings suggest that in terms of consumer review attitude ( Ar ), affect is a more direct an d dominant predictor, whereas cognition seems to play a more critical and immediate role in forming product attitude ( Ap ) and purchase intention ( PI ). T he study also found a dominant moderating effect from valence of consumer review whereas the effects fro m the other moderating variables product types and involvement levels were merely detected. In positive consumer review situation, cognition dominated affect for predicting consumer review attitude ( Ar ), product attitude ( Ap ), and purchase intention ( P I ) regardless of product type and involvement level. On the other hand, when people were exposed to negative consumer review situation, the results were totally the opposite, i.e., affect, in turn, dominated over cognition for predicting consumer review at titude ( Ar ), product attitude ( Ap ), and purchase intention ( PI ) regardless of product type and involvement level The theoretical and practical implications of these findings are discussed.
11 CHAPTER 1 INTRODUCTION People want to make good dec ision s People who attempt to make good decision s will often refer to the opinions of others to help make up their own mind; this is even truer when making a decision as a consumer who is attempting to reduce risk (Hennig Thurau & Walsh, 2003). Consumers r educe risk by seeking out information and being influenced by advertising, publicity, salespeople, peers, the Internet, and TV news. While there are numerous information sources available, consumers are likely to gather third party opinions when making dec isions (Wang, 2005). These third parties are considered non marketer dominated sources and include critiques, peers, and word of mouth (WOM) referrals. Such sources are not supposed to have a personal stake in as more credible and less biased (Hoyer & MacInnis, 2007). Among those sources, WOM is widely available and considered to be a critical component of marketing in recent years because consumers often seek out WOM opinions before they purchase books, movie tickets, IT products, cars, or choose restaurants. Consequently, WOM is generating an increased interest from marketers (Chevalier & Mayzlin, 2006). The Rise of WOM in the Internet Era Word of mouth is generally defined as interpersonal communication wit h a verbal exchange of positive and negative information about products and services (Haywood, 1989). Many researchers have shown that WOM is one of the most influential marketing elements (Bayus, 1985; Richins, 1983; Rosen, 2000; Whyte, 1954). Katz and La zarsfeld (1955) found in their classic work that the influence of WOM is twice as important as personal selling and 7 times more important than print advertising to consumers making
12 purchas ing decision s Arndt (1967) also maintains that WOM effects consume decision making in a wide range of product categories. In recent years, as the Internet has become a revolutionary phenomenon of communication, WOM has naturally incorporated itself with the Internet, and, in so doing, it has become empowered even mor e (Hennig Thurau et al., 2004). Today, companies take online WOM (eWOM hereafter) as one of the most powerful marketing forces and opportunities since consumers now more than ever tend to pay attention to eWOM. Although there are a number of ways in wh ich eWOM messages are communicated through the Internet, online consumer review s are the most common form of eWOM that is readily available and frequently accessed by consumers (Hennig Thurau et al., 2004; Sen & Lerman, 2007). As online shopping has become a prevalent often make online purchases after they have consulted online consumer reviews provided by the web sites for books ( e.g. Barnesandnoble.com), travel ( e.g. Pric eline.com), or just about any product from MP3 players to home improvement items ( e.g. Amazon.com). Because of the nature of online shopping where online shoppers cannot see and feel the goods personally as they can in brick and mortar stores, more consum ers are making their purchase decisions based on online consumer reviews that provide indirect experience s with the goods from other consumers (Park, Lee & Han, 2007). Also, Epinions.com, Consumerreview.com, Rottentomatoes.com, or Kellybluebook.com are on ly some of the numerous popular web sites that are not online shopping malls but rather provide platforms for consumer reviews to share experiences
13 and opinions about products and services. According to recent research conducted by Forrester.com, more than when they make purchase decisions (2005). Consequently, the importance of online consumer review is increasing for marketers (Park, Lee & Han, 2007). Need for the Study Although online consumer re view s (or eWOM) ha ve received substantial coverage by the trade and popular press for many years, little formal research had been published in scholarly journals until recent years (Park & Chung, 2006; Rosen, 2000; Sen & Lerman, 2007). The studies that hav e been published in scholarly journals mainly focus on user motivation or the effect aspect of eWOM (Bickart & Schindler, 2001; Chatterjee, 2001; Chen & Xie, 2008; Cheung, Lee & Rabjohn, 2008; Gruen, Osmonbekoy & Czaplewski, 2006; Hennig Thurau & Walsh, 2 003; Hu, Liu & Zhang, 2008; Samson, 2006). Despite the significance of online consumer review s as a form of eWOM communication and as a critical factor in marketing, there are still many questions yet to be answered, and many areas yet to be explored. One of the critical voids understudied in the arena of the online consumer review is probably the process of how consumers form their attitudes towards a brand/product based on online consumer reviews. ume communication. A s fast as continuously create new ways to approach consumers and develop new marketing communications. been e xploring new and innovative methods to exercise control over WOM (Breazeale, 2009) which is supposed to be a non marketer dominated form of communication
14 has become one of the m Despite the use of this term, less attention has been focused on the underlying mechanism of the process of how consumers are influenced by online consumer reviews, one of the most common types of eWOM, in forming attitudes towards a brand/product and finally in m aking purchas ing decision s The purpose of the current study is two fold. First, Main Study I will examine formed by their thinking about the product/brand under various conditions (e.g. product type, level of involvement, and valence of eWOM messages). This study will examine the effects of those various conditions on affective and cognitive based attitude s toward online consumer review s toward brand s and ultimately decision making. Second, Main S t udy II will investigate how affective and cognitive processing two of the main components in forming an attitude play roles in the impact of online consumer reviews on brand/product attitu de formation, and consequently on purchase intention formation The study will attempt to develop and test a conceptual model to explain this process by using the Structural Equation Modeling (SEM) technique. Attitudes have become one of the most popular and essential constructs for most social science domains today (Eagly & Chaiken, 1993; p.1), especially marketing because marketing research in attitudes ( e.g., attitudes towards marketing communications or brands/products) provides marketers with critic al information to develop marketing strategies and tactics that connect them with consumers. Understanding how consumers form attitudes also enables marketers to predict consumer behavior
15 ( purchasing intention of products or services ) which is a critical part of the goals of marketing. By achieving the two goals, the current study is expected to cont ribute to understanding the underlying mechanism of how consumers are influenced by online consumer reviews, one of the most common, and powerful eWOM phenomen a in the marketing process today.
16 CHAPTER 2 LITERATURE REVIEW AND HYPOTHESES Class ic Studies in WOM in Fortune WOM has been defined by several researchers (Aaker & Myers, 1982; Bayus, 1985; Haywood, 1989). These traditional definitions share a common ground in that WOM is an exchange of information by verbal means in an informal, person to person manner. Katz and Lazarsfeld (1955) were pioneers of survey analysis and con ducted the empirical study of WOM. In their classical personal influence study in 1955, they determined how individuals obtained information and opinions from others to make their decisions. The study interview e d a group of more than 800 women in Illinois to examine the influence of opinion leaders on general consumers relative to mass media and other marketing communication sources. Katz and Lazarsfeld suggested a two step flow model as a WOM process (Figure 2 1). The model suggested that, unlike the tradi tional model, there are opinion leaders between the mass media and general consumers (followers), and those opinion leaders have a greater and more direct influence on general consumers than the mass media. Arndt and May (1981) hypothesized a dominance hi erarchy of information sources, which maintains the existence of a direct hierarchy of influence among different types of sources (Figure 2 2). According to Arndt and May, the use of WOM communication (interpersonal sources) depends on the level of brand e xperience (direct prior experience), whereas the use of advertising (mass media) depends on the level of WOM information. Based on a logical process of deduction, interpersonal sources rank
17 lower in perceived usefulness than direct prior experience but ran k higher than mass media. Therefore, direct prior experience (brand experience) tends to dominate interpersonal sources (WOM), and interpersonal sources tend to dominat e mass media (advertising). Th e researchers supported this notion by comparing and contr asting the structural and operational characteristics of these three different source levels, including attribution of biases, opportunity for feedback, control of feedback, relevance of content, completeness, validity, and accuracy. Although this idea was originally developed for durable consumer goods, Faber hypothesis to movie most cases, people are influenced by mult iple sources that can provide conflicting information regarding a new movie. The receivers in this situation must eliminate the contradiction between sources and reach agreed with Chaffee (1979) on the notion that peo ple learn source credibility by using and comparing different information sources through repeated experiences. Over time, people gradually perceive some sources as more credible than others. Given this erceptions of different consultation, perceived credibility, importance, and usefulness: one direct prior ads, magazines), and three interpersonal sources (comments from friends, comments from a spouse/date, comments from someone known by the res pondent and considered
18 to be a movie expert). The authors found that interpersonal sources were generally more influential than mass media sources in selecting movies. More recently, Acland (2003) completed another study that supports Arndt and going decision making by surveying college students. Students were asked what attracted them to the movies they had most recently seen. The students w ere given twenty eight options to choose from to rate the importance of information sources that could be used to make the i r decisions for the s (WOM) w ere rated as the second most important influenc e after to movie characteristics ( e.g., plot and genre) and followed by mass media advertising, previews, etc. eWOM One of the unique characteristics of the Internet is its interactivity. The emergence and prevalence of the Internet make it possible for co nsumers to interactively share their thoughts and experiences about products, services, and brands with other people more easily than ever (Schindler & Bickart, 2005). This kind of information exchange, in other words, WOM, which is generated on the Intern & Bickart, Schindler and Bickart (2005) claim that there are a number of ways in which WOM messages are communicated through the Internet, and they can be divided into seven review (referre d to type the current
19 study analyzes because it is currently considered the most commonly found form of that includes co nsumer and reader comments and feedback posted on the web sites of time conversations between groups of people over the Internet take place. Final on one real time conversations over the Internet (Schindler & Bickart, 2005). Goldsmith and Horowitz (2006) noted in their research that eWOM is an important aspect of ecommerce. According to the authors, eWOM affects t he sales of products and services because consumers tend to actively give and seek opinions online in the same manner that opinions are traded offline. According to Hung and Li (2007), eWOM could be considered even more effective because it provides explic it information, tailored solutions, interactivity, and empathetic listening directly to consumers. Also, several other studies (Bang, 2006; Hennig Thurau et al., 2004; Jung, 2006) confirm that eWOM could be more powerful in communication than traditional W OM due to its distinct characteristics ( e.g., availability to reach unspecified multiple individuals with no geographical/time limit) and the impressive technological development of the Internet ( e.g., enhanced interactivity, information searchability ). Di stinct Characteristics of eWOM versus WOM There are several known important differences between traditional WOM and eWOM. First, consumers are no longer constrained by time, place, or acquaintances
20 either in transmitting or receiving information with eWOM as they are with WOM. That is, traditional WOM is typically conducted face to face, whereas eWOM is web based communication that overcomes most of the physical barriers that hold back traditional communication (Hennig Thurau et al., 2004). Second, the amou nt of information and the number of sources that consumers can access online is greater than what is available offline (Chatterjee, 2001). Offline, WOM is limited to those a consumer can physically contact. Through the Internet, however, consumers have acc ess to a larger and more diverse set of opinions about products and services posted by individuals who have used the product or are knowledgeable about the service, yet consumers need not have a prior relationship with those individuals to take advantage o f the information provided (Schindler & Bickart, 2005). Third, eWOM enhances the cost efficiency of acquiring information. It saves time, effort, and money to find the appropriate information compared to searching offline in a more traditional way (Bang, 2006). All these traits give eWOM a far greater reach than most traditional offline methods of gathering information. eWOM and Attitude As previously discussed, there ha s not been not m uch academic research on how people form their attitude s when respondin g to eWOM, despite it being one of the most intriguing topics in mar keting communication today ( T he concept of attitude will be detailed in the next section.) In fact, this area of study has been pioneered and led by only a handful of foreign scholars. Pa rk, Lee and Han (2007) investigate the effects of focus on how the quantity of consumer reviews (the number of online consumer reviews), and quality of consumer reviews ( the number of arguments presented in a
21 consumer review) influence purchase intention of the product. Their findings revealed positive correlations between purchase intention both with quantity and quality of s purchase intention s in a low involvement situation tend to be more influenced by the quantity rather than the quality of consumer reviews, which is in line with the findings of Elaboration Likelihood Model (ELM) ; h owever, consumers in a high involvement situation tend to be more affected by the quantity of consumer reviews only when the quality of the reviews is high. Lee, Park and Han (2007) conducted another interesting investigation that product attitude s, especially when the valence of the review is negative. The study detected a strong influence of negative consumer reviews on consumer attitude especially when the quality of the review ( refe r r ing to the number of arguments presented in consumer review) is high. More s pecifically, consumers in a low involvement situation tend to be more easily influenced (i.e. view a brand as unfavorable) by negative consumer reviews whereas consumers in a high involvement situation tend to show the grea ter conformity effect only when the quality of the negative reviews is high. Park and Lee (2008), this time, control the valence of consumer reviews as positive and examine the effect of eWOM in a very practical setting ehavioral intention. eWOM overload (or information overload: e.g., too many consumer reviews for a certain product) happens when there is too much information about a product/service provided to consumers, which may have a negative nitive processing, their attitude toward a brand and may eventually work adversely for marketers. The researchers detected a negative effect of
22 even if positive message s are included T he researchers however, also found that in a low involvement situation, eWOM overload may generate perceived popularity among consumers and eventually increase the purchase intention. The Concept of Attitude Attitude is the most importa nt concept to psychological tendency that is expressed by evaluating a particular entity with some & Chaiken, 1993; p.1). Generally, attitude is about how someone views or evaluates a person, object, action, issue or a thing (Petty, Unnava & Strathman, 1991). Attitudes are considered a state of mind that can be changed by persuasion. Attitudes change as a response to communication ; t herefore, over the past century, the concept of at titude has become one of the most important areas of interest of scholars in social science including advertising, marketing and communication (Perloff, 2008). Especially in advertising and marketing communication perceived as a type of persuasive commun ication, attitude toward advertising or brand has been one of the most frequently measured indicators to evaluate the effectiveness of communication (Jun, Cho & Kwon, 2008). The concept of attitude is a very complex, hypothetical construct. It cannot be o 2008). Researchers have investigated the structure of attitude for many decades result ing in controversies, mixed use of terms, and conflicting findings
23 Tripartite At titudinal Structure Although findings of attitude research have been confusing and indecisive, most scholars in the area seem to agree that attitude is an inferred state that can be divided into three classes: cognitive, affective, and conative (behavior) processes (defined below). This multidimensional approach of attitude has been widely accepted for many decades of attitude study (Bagozzi & Burnkrant, 1979; Breckler, 1984; Brown & Stayman, 1992; Eagly & Chaiken, 1993; Holbrook & Batra, 1987; Katz & Stotl and, 1959; Zajonc & Markus, 1982). Among th e three components, scholars have long been seeking better understanding of the roles of affective and cognitive components. A ffect ive and cogniti ve components are thought to last component behavioral intentions (Bagozzi & Burnkrant, 1979; Morris et al 2002). Katz and Stotland (1959) maintain that all true attitudes comprise cognitive and affective content. Rosenberg (1968) adds : T he authors and proponents of most of the other consistency theories pay ready lip service to the definition of attitude as an internally consistent structure of affective, cognitive and behavioral components. But, in practice, the last of these components is usually slighted. Behavior (in the sen se of externally visible, overt action) toward the attitude object is usually relegated to the status of a dependent variable; implicitly it is assumed that the person will simply act toward an attitude object in a manner consistent with his coordinated af fective cognitive orientat ion toward that object (p.101) The majority of attitude studies ha ve been focusing on affective and cognitive components of attitude and treating the behavioral (conative) component of attitude as dependent effects of cognitive and affective variables (Bagozzi & Burnkrant,1979). Since Bagozzi and Burnkrant (1979), cognition and affect the two dimensions of
24 attitude construct have been consistently studied by advertising and marketing communications researchers (Jun, Cho & Kw on, 2008). Cognition Cognitive response includes thoughts, learning, knowledge, or ideas about the attitude object that are often conceptualized as beliefs ( e.g., & Chaiken, 1993; Fis hbein, 1963). Beliefs often play a critical role in creating associations between the attitude object and its attributes (Fishbein & Ajzen, 1975). Cognition is also conceptualized as the int ernal representation of reality that is organized and acquired b experience s (Buck, 1988). Affect Affective response is feelings, moods, or emotions that people experience when responding to an attitude object ( e.g., & Chaiken, 1993). It is generally perceived that affect, emotions and feelings are interchangeable terms in most advert ising literatures (Homer, 2006) whereas mood is generally considered a state of mind that is longer lastin g and less intense than affect, emotions or feelings. M ore importantly, moods generally have no direct association with any particular stimulus object, which differentiates moods from affect, emotions, and feelings (Hansen & Christensen, 2007). The three dimensional view of affect Evaluation, Activation and Surgency (Osgood, Suci & Tannenbaum, 1957) more recently described as Pleasure, Arou sal, and Dominance (PAD) theory is one of the most widely accepted and studied theories regarding affect (Lang, 1995; Osgood, Suci & Tannenbaum, 1957; Russell &
25 Mehrabian, 1977). According to the PAD theory, affect is formed on the basis of three independent and bipolar dimensions : pleasure displeasure, degree of arousal (excited calm), and dominance submissiveness People are in a constant state of emotion within a three dimensional space of P, A and D (Russell & Mehrabian, 1977). The theory suggests that any type of affect (emotion) can be understood as a unique combination of the three primary dimensions: P, A, and D ( e.g., pleasure and arousal, b ut with low dominance) (Michael & Morris, 2004). Because of the usefulness and reliability of the thre e dimension al theory as a model for measuring human emotions, the theory has gained widespread attention from scholars in advertising and marketing research (Havlena & Holbrook, 1986; Zeitlin & Westwood, 1986). There have been a lot of measurements explor ed and developed by researchers to measure emotions. While interview s (Richins, 1997), experimental manipulation (Shiv & Fedorikhin, 1999), psycho physiological measures (Aaker et al 1986), and facial expressions (Verma et al 2005) were some of the att empts (Hansen & Christensen, 2007), AdS AM the Attitude Self Assessment Manikin by Morris (1995) is perhaps one of the most practical, useful and functional measurements for affective response especially to marketing communication stimuli (Morris et al 2002). AdS AM uses non Assessment Manikin) and the three dimensional PAD model by Russell and Mehrabian (1977).
26 AdS AM uses three sets of visualized graphic characters arrayed along with a continuous 9 point scale to represent the three dimensions of affective responses to stimuli pleasure, arousal, and dominance without interference of cognitive processing (Mehrabian & Russell, 1974; Morris et al 2002). The first row of figures represents the pleasure scale (P), ranging from pleasure to displeasure ; the second row represents the arousal scale (A ), ranging from excited to calm ; and the third row represents the dominance scale (D), ranging from controlled to controll ing. Morris and his colleagues investigate the human brain in order to validate the three dimensional construct of emotion. In a recent study by Morris et al. (2008), emotional responses to five video clips (in this case television commercials) were exami ned with AdS AM and functional magnetic resonance imaging (fMRI) to identify corresponding patterns of brain activation. Significant differences were found in the AdS AM scores on the pleasure and arousal rating scales, which suggest a dimensional approach of constructing emotional changes in the brain. AdS AM has been one of the most effective measuring tools for affective responses and is applicable to various marketing communications (Morris, 1995; Morris et al., 2002; Morris et al., 2008). It has been u sed in many different studies to measure affective responses to a variety of stimuli including product concepts, finished ad s product attributes, product benefits, brands, logos, tag lines, packaging, music, etc. (Jin, 2006). This is d iscussed further in the method section Conation Conative (behavioral) response is a predisposition toward action (Traindis, 1971). It can be regarded as both the overt actions that people exhibit in responding to an attitude object (e.g. circulating petitions opposing nuc lear power plant construction),
27 and behavioral intentions that are not necessarily exhibited in overt actions ( e.g., intention to circulate petitions tomorrow, regardless of whether the intention results in action ) (Eagly & Chaiken, 1993). Conative respons e of attitude is generally measured by the semantic differential scales, and refers to how cognitive and affective components are associated with their behavioral intention to decide what to do (Kane, 1985; Mischel 1996). This is disc ussed further in the following section Attitude Formation: Cognition and Affect As discussed above, among th e three components cognition, affect, and conation affect and cognition have been identified a s two antecedents of attitude by many researchers ( e.g., Bagozzi et al., 1999; Ervelles, 1998; Holbrook & Hirschman, 1982). Holbrook (1978) explains that cognition deals with logical, objectively verifiable descriptions of the tangible features of an attitude object, whereas affect comprises emotional, s ubjective impressi ons of intangible aspects of an attitude object. Holbrook was one of the scholars who stressed the importance of both components in shaping s toward a certain object in his two dimensional attitude construct theory (Holbr ook & Hirschman, 1982). J ust a few decades ago, in the early stage s of consumer behavior research, the cognitive based approach was a dominant research trend and scholars seemed to perceive consumer choice behavior as merely information processing (Hansen & Christensen, 2007; Morris et al 2002). For a long time, the affective component of attitude formation was largely neglected and did in social science. Affect was often regarded as disruptive and disorganized beh avior, and a primary source of human problems. Consequently, research on the role of affect
28 in forming attitudes seemed to be unscientific until recent years (Izard, 1991 ; Morris et al 2002). Today, despite the widely accepted theory of attitude formati on that suggests the roles of affect and cognition, many questions still remain unanswered including in quiries regarding how affect and cognition, in practice, interplay with each other in shaping attitudes, and wh ich is more effective affect or cognitio n in predicting behavioral int ention Interplay between Cognition and Affect As previously discussed, until recent decades, the affective component of attitude was largely ignored and underestimated by cognitive oriented scholars (Morris et al 2002). Most researchers focused on the cognitive component of attitude suggesting et al 2002). It was widely perceived that the role of affect was limited, and any influen ce it held on attitude t ook place through cognition as a mediator. Cognition was perceived to be indispensable to affect in forming attitude, and affect was thought to occur only as a result of the cognitive process (Lazarus, 1982; Lazarus, 1984; Tsal, 198 5). Specifically in marketing communication, Fishbein (1965) was another contributor in th e notion that suggest ed primarily a function of cognition. Zajonc (1980) was one of the early challengers to the traditional notion of cogniti ve based attitude. He agreed that in some cases, the cognitive component may be dominant, and attitudes may be formed solely based on cognition with no affective components involved (e.g. A favorable report on an unknown product from an authority can crea te a favorable attitude towa rd the product for some people. ). However, he
29 maintained that, under certain circumstances, affect could be evoked prior to cognition, and therefore may at times play a more critical role than cognition in shaping attitudes. S ince then, researchers began to realize the importance of the role of affect in forming attitudes that had long been underestimated and to explore the interrelationship between cognition and affect in shaping attitudes du ring the communication process Esp ecially in marketing communicati ons, a number of researchers ha ve been studying to acquire a better understanding regarding the interplay of cognition and affect. Homer (2006) is one of the researchers who argued that the role of cognition in attitude for mation was overestimated. In contrast, her study found that affect exerted more dominant influence on the attitude formation process when the brand is unknown than when the brand is well known. Leigh et al. (2006) used print ads to investigate whether two dimensions of memory (recall and recognition) for print ads are associated with cognition and affect. It was found that recall is more influenced by cognition than affect, whereas recognition is more influenced by affect than cognition. Morris et al. (200 2) attempt to extend the research scope and examine how affect and cognition can influence attitude and even predict behavioral intention. The researchers assert that, if measured more rigorously with no interference of cognitive processing in measuring af fect by using AdS AM a non verbal measure of affect, explaining the variance toward behavioral intenti on. The relationship among attitude, behavioral intention, and the ir antecedents affect and cognition is discussed more in the following section
30 Some researchers even noted the independent role of affect without the presence of cognitive proc ess. The mere exposure theory is one of the empirically tested and widely accepted notions that prove the influence of affect imposed directly to attitude (Clore & Schnall, 2005; Zajonc, 1968). The theory suggests that a stimulus in a certain situation (e.g. repeated exposure to attitude objects or to information about them) m ay directly elicit affect (e.g. favorability toward the stimulus) without any cognitive mediation (Eagly & Chaiken, 1993; Zajonc, 1968; Zajonc, 1980). Today, most researchers agree and accept the notion that attitude is dev eloped and formed in various combinations of cognition and affect (Perloff, 2008). Based on the previous research, the following hypotheses are submitted: H 1 : T he effect of affective response toward consumer reviews will have a positive effect on consumer review a ttitude H 2 : T he effect of cognitive response toward consumer reviews will have a positive effect on consumer review a ttitude Affect and Cognition: Predicting Behavioral Intention Can attitude predict behavioral intention? This current study al so looks at how cognitive and affective based attitude s toward consumer reviews can predict consumer behavior purchasing intention of products or services which is a critical issue in marketing research. Traditionally, the power to predict human beha vior has been a central interest of any social science ( e.g.,
31 the issue of attitude behavior consistency to p redict behavior from attitude has been one of the most important questions for researchers to explore (Morris et al 2002). However, because of the unsuccessful investigations of early researchers (Corey, 1937; LaPiere, 1934) in finding the capabili ty of attitude to forecast behavior, were skeptical until the 1970s and 80s (Perloff, 2008). O ne such unsuccessful investigation is the classic work of LaPiere (1934) at Stanford that empi rically tested the relationship between attitudes and behavior and showed a huge gap between the two constructs. LaPiere s (1934) study was a seminal study in establishing the notion that behavior may not be predictable based on verbally reported attitudes More rigorous research methods, however, changed the notion of scholars in the 1970s. Drawing on the previous findings in social psychology, Fishbein and Ajzen (1975) formulated the Theory of Reasoned Action (TRA), a well established theory that provide s numerous social science studies with a framework to specify the impact of attitudes on, and therefore, to predict behavior (more specifically behavioral intention rather than overt behavior) (Hoyer & MacInnis, 2007; Perloff, 2008). According to TRA, beh avioral intention is a function of two frequently conflicting determinants, attitude toward the behavior ( e.g., the subjective norm (social pressure, e.g., The intention itself serves as the immediate determinant of behavior. Attitude to behavior theory also suggests that attitudes have direct influence on behavior (Fazio, 1986). The evidence of the relationship between attitudes and behavioral intention (in most cases, pu rchase intention), e specially in numerous
32 advertising and marketing communication studies, has been well documented and confirmed (Jun, Cho & Kwon, 2008; MacKenzie, Lutz & Belch, 1986; Muehling, 1987; Shimp, 1981). Researchers have found that attitudes t oward ads ( Aad ) often have a direct influence on brand attitudes ( Ab ) as the favorable (or unfavorable) attitudes toward the ads are often transferred to the advertised brand, and eventually increase (or decrease) the purchase intention ( PI ). N o research h owever, has been conducted to investigate how attitudes influence the behavioral intention in the context of consumer review s to the Because of this shortcoming, t he current study extends the discussion to the relationship among a ttitudes toward consumer reviews, attitudes toward a product/brand, and purchase intention. Based on the extant literature on the relationship between attitude and behavioral intention, the following hypotheses are submitted: H 3 : Consumer review a ttitude w ill have a positive effect on brand/product attitude H 4 : Brand/product a ttitude will have a positive effect on purchase intention. Moreover, if it is certain that attitude has an impact on behavioral intention, the current study pursues the inquiries furt her into the issue about what specific component of attitude (i.e. cognitive or affective) is more predictive of behavioral intention under various conditions in the context of a consumer review by testing the overall hypothesized model (Figure 2 4) using SEM analysis. Attitude Formation under Various Conditions The different roles of affect and cognition in the evaluation process have been investigated under various conditions and in different contexts by a number of scholars. Most frequently, attitude fo rmation processes have been compared across types of
33 product s (hedonic vs. functional) in marketing literature because different types of differently (Batra & Ahtola, 1990). Hedonic products are t he types of products consumed primarily to pursue affective benefits (e.g. taste of toothpaste), whereas functional (a.k.a. utilitarian) products are c onsumed to seek more cognitive oriented benefits (e.g. tooth decay) (Ba tra & Ahtola, 1990; Kempf, 1999; Woods, 1960). Researchers have found that different product attributes contribute differentially to the two different product types and attitude formation (Batra & Ahtola, 1990). For hedonic products, presumably affective p rocessing of product information (consumer review s ) may be more influential, whereas for functional products, cognitive processing may be more influential. Another variable that may influence attitude formation is the level of involvement (situational inv olvement is the focus for the current study). Researchers have found the significant moderate effect of consumer involvement in attitude formation in the context of product trial experience. One of the early works was don e by Batra and Stephens (1994) who found that the role of affect and cognition in attitude formation may vary depending on conditions ; however, the involvement level i s more important than product type in predicting the role of cognition and affect. The researchers found that the role of af fective responses were more critical in shaping brand attitudes in low involvement situation s whereas cognition played a more critical role in high involvement situations. Kempf (1999), however, finds somewhat conflicting findings in his experiment where he detects the moderating effect of product types. According to his study, attitudes are influenced by both affect and cognition ; however, for hedonic products,
34 affect is more important in forming attitude, whereas for functional products, cognition may pl ay more important role. Kim and Morris (2007) also investigate how affective and cognitive response s to a product trial experience exerts influence on attitude formation of the product under different product type s (hedonic vs. functional) and involvement (high vs. low) conditions. The researchers found the dominant power of affect in forming an attitude toward a product trial, whereas the influence of cognition and affect on product trial based attitude toward product appears fairly balanced. The current s tudy is greatly inspired by Kim and Morris (2007), and adopts some of the ir methodological framework the current eWOM study. S imilar research conducted by Pham et al. ( 2001), however, finds that the affective response may be a more dominant predictor than cog nition regardless of the type o r combination of conditions. Based on the extant literature on the role of cognition and affect in shaping attitudes under various con ditions, the following research questions are submitted: RQ1: How differently would the affective and cognitive response s toward cons umer reviews influence consumer s attitude formation toward consumer reviews under the different product type / involvemen t situations? H 5 : Under the functional product in a high involvement situation, the effect of cognitive response toward consumer reviews on consumer review attitude formation will be greater than affective response. H 6 : Under the functional product in a low involvement situation, the effect of cognitive response toward consumer reviews on consumer review attitude formation will be greater than affective response.
35 H 7 : Under the hedonic product in a high involvement situation, the effect of affective resp onse toward consumer reviews on consumer review attitude formation will be greater than cognitive response. H 8 : Under the hedonic product in a low involvement situation, the effect of affective response toward consumer reviews on consumer review attitude formation will be greater than cognitive response. RQ 2 : How differently would the affective and cognitive response s toward consumer reviews influence consumers attitude formation toward product s under the different product type / involvement situations? H 9 : Under the functional product in a high involvement situation, the effect of cognitive response towa rd consumer reviews on consumer s product attitude formation will be greater than affective response. H 10 : Under the functional product in a low invo lvement situation, the effect of cognitive response toward consumer reviews on consume r s product attitude formation will be greater than affective response. H 11 : Under the hedonic product in a high involvement situation, the effect of affective response towa rd consumer reviews on consumer s product attitude formation will be greater than cognitive response. H 1 2 : Under the hedonic product in a low involvement situation, the effect of affective response towa rd consumer reviews on consumer s product attitu de formation will be greater than cognitive response. RQ 3 : How differently would the affective and cognitive response s toward cons umer reviews influence consumer s purchase intention formation under the different product type / involvement situations? H 1 3 : Under the functional product in a high involvement situation, the effect of cognitive response towa rd consumer reviews on consumer s purchase intention formation will be greater than affective response.
36 H 14 : Under the functional product in a low invol vement situation, the effect of cognitive response towa rd consumer reviews on consumer s purchase intention formation will be greater than affective response. H 15 : Under the hedonic product in a high involvement situation, the effect of affective response towa rd consumer reviews on consumer s purchase intention formation will be greater than cognitive response. H 16 : Under the hedonic product in a low involvement situation, the effect of affective response towa rd consumer reviews on consumer s purchase in tention formation will be greater than cognitive response. A c onsumer review as a word of mouth contains information about products or services, and it is either positive or negative in its nature (Haywood, 1989). T he valence of th is information may exert critical influence on consumers attitude formation and generate critical difference s in results. In fact, researchers have been debating about the effectiveness of positively versus negatively framed messages under high versus low cognitive elaboration situation (Lee, Park & Han, 2007; Shiv, Britton & Payne, 2004). The following research questions are submitted: RQ 4 : How differently would the affective and cognitive response toward consumer attitude formation (toward consume r review/product) and purchase intention formation under the different valence situations? Figure 2 1 The Two step Flow Model [Adapted from Katz and Lazarsfeld 1955 ]
37 Figure 2 2 The Hierarchy of Information Sources [Adapted from Arndt and May 1981 ] Figure 2 3 AdS AM (Self Assessment Manikin) [Adapted from Morris 1995 ]
38 Figure 2 4 Hypothesized model and hypotheses
39 CHAPTER 3 METHODS Overall Procedure The purpose of the cur rent study is to cognitive responses toward online consum er reviews when thinking about the product/brand under various conditions of product type, level of involvement, and the valence of the information in the consumer reviews (Main Study I) T herefore, t he current study was conducted under 2 x 2 x 2 experimen tally designed conditions with manipulated valence (positive vs. negative), product type (hedonic vs. functional), and involvement level (high vs. low) In order to manipulate the conditions t hree pretests w ere conducted prior to the main experiment. The entire processes of the three pretests are discussed in detail in the later part of Chapter 3. The current study also aims at develop ing and test ing a conceptual model to explain how two of the main components in forming attitude affect and cognition p lay their roles in the impact of online consumer reviews on brand/product attitude formation, and consequently purchase intention formation (Main Study II) Experiment wa s adopted for the current study b ecause one of the main goals is to examine the caus al relationship among constructs in order to understand how affect under various conditions. E xperiment is arguably the best research method for social sciences to establish cau se and effect for several reasons : ( 1) i t enables researchers to control the order of the presentation between the cause and the effect ; ( 2) it helps researchers leave out extraneous variables ; and ( 3) it helps researchers to t ake control over
40 environments that may affect the results. Th e se issues are very critical conditions for determining causality (Wimmer & Dominick, 1997). Experiment subjects for the pretests and the main experiment w ere college students in the United States. Subjects for the prete sts ( n=39, 20 males and 19 females ) were undergraduate students recruited from summer classes in a large southeastern university in July, 2011 All of the subjects voluntarily participated in the study and earned some extra credits for their summer classes S ubjects for the main experiment (n=250, 121 males and 129 females, age range 18 to 29) were also undergraduate students recruit ed through a mark e t research panel company, uSamp in February, 2012 There was no direct compensatio n from the researcher for research participation. Th e convenient sampling method of using only a homogeneous group such as college students may limit the external validity (generalizability to the general population) of the study ; h owever, using a homogenous group with high commo nality may be also advantageous for experiment studies like the current one to examine a theoretical relationship between constructs because it enables the researcher s to maintain more control over the sample throughout the entire study. Also, as argued by Kempf (1999), the results may yield more powerful implications for a specific demographic group because college students may be one of the most important target groups for certain product/service markets. P rior to start of the main experiment, each parti cipant was asked to answer a question to show their level of interest in the product to be used for the main experiment. Data from those who show ed no interest, or significantly lower interest than
41 other participants w ere filtered out in order to control any possible extraneous variables or outliers. T he detailed process es and results of each pretest are explained below. Pretest 1 Manipulation 1 (Product Type) The goal of P retest 1 wa s to select a pair of product s (product category) to be used for the ma in experiment The product category selected by Pretest 1 has to be indeed relevant to the specific demographic group participating in the current study college students. Also, t he t wo products in the pair should be different from each other in terms of their hedonic and functional nature, but similar in all other aspects ( e.g., a grammar boo k vs. a cartoon book ) A small interview (n=10 4 males and 6 females ) and a paper and pencil survey (n=39) were conducted for Pretest 1. The purpose of the intervie w was to brainstorm and select a pair of hedonic and functional goods that are relevant for college students The interview was moderated by a third person a trained graduate student in mass communication who wa s not informed of the purpose of the study At the beginning of the interview for brainstorming the definitions of hedonic and functional goods (as defined in Chapter 2 ) w ere given to the 10 participants who were then asked t o freely talk about their ideas T he question was worded Based on the definitions of hedonic and functional goods we just discussed, please give examples of pairs of products that are different from each other in terms of their hedonic and functional nature, but similar in all other aspects The moderator wr o te down the ideas about product selection At the end of the brainstorming, a list with 3 pairs of products w as s Products that we re considered to be inappropriate for the current study were excluded from the list. The final list of the 3 product pairs was the n evaluated by a separate group
42 of participants (n= 39 20 males and 19 females), which wa s another set of college students through a paper and pencil survey. The p articipants evaluate d each pair of hedonic / functional products on a 9 point scale by answe ring the following question below as executed by Kim and Morris (2007) : Would you characterize this product as primarily a functional product or an entertain ment/enjoyable product? P rimarily for functional use, P rimarily for entertainment use A paired t test analysis was conducted to examine whether there wa s a statistical mean difference between functional and hedonic products for each pair as shown in Table 3 1 While all of the 3 pairs showed statistically significant mean differences, Pair 3 (grammar software vs. game software) indicate d the most critical mean difference (hedonic product score minus functional product score) and was therefore selected for the main experiment. As a manipulation check between grammar software vs. game sof tware, the same analysis was repeatedly conducted with a larger sample (n=250) of the main experiment During Pretest 1, participants were asked to write down the attributes that they thought to be important for each product. Those salient attributes of ea ch product were collected to develop the measurement for one of the constructs, c ognitive response toward c onsumer r eviews As explained later in Chapter 3 expectancy value measures ( ), which consist of attribute level brand beliefs (Bi) and attribute evaluations (Ei) were adopted to measure the construct. The four most indicated salient attributes for the pair of product types chosen by Pretest 1 for the main experiment (gramma r software vs. game software) are shown in Table 3 2.
43 Pretest 2 Manipulation 2 (Involvement) P retest 2 w as conducted to pre check the manipulation of involvement to be used in the main experiment. Involvement is generally defined as the extent to which a n individual consumer personally perceive s relevant to an entity (Krugman, 1966). Among several types of involvement ( e.g., message, product, situation, etc.), situational involvement w as used because it can better explain the relationship between consumer & Morris, 2007). Situational involvement refers to the involvement evoked by a specific situation and one of the most widely used measures in advertising and marketing research is the purchase decision situation (Kim & Mor ris, 2007), which i s explained below and used for the current study as well. For the current study, two different hypothetical purchase situation scenarios w ere adopted from Kim and Morris (2007) and modified for high vs. low involvement condition manipul ation as below. High involvement situation went to a store to buy grammar check ing (or game) software a birthday present for my lovely brother who wants to be a professional writer (or gamer) I have been searching for software that my brother would really want, and I have finally found one that he would like. This product is my final choice among many seen in the local stores and on the Internet. Because his birthday party is tomorrow, I need t o buy this one, if it seems to be a good fit. The price of the software is $49.99. I need to be really careful in making my decision because the store provides no return Low involvement situation I just saw grammar check ing (or game) software running on a dis play monitor while I w as shopping for other products. The price of the software is $49.99. The store provides a generous 60 day no questions asked return
44 policy for this product. I am a little interested in this grammar check ing (or game) For P retest 2, the study followed up with the same group of pa rticipant s (n= 39 20 males and 19 females) used for Pretest 1. The participants were asked to answer a 3 item 9 point semantic differential scale survey. This process wa s to evaluate if the difference between a manipulated high involvement condition and a low involvement condition wa s significant enough for the current study. The questions that were used are below as de veloped by Mittal (1989) : Based on the situation you were given, in selecting thi s product from many other choices availabl e in the market, would you say? Based on the situation you were given, how important would it be for you to make a choice on this product? (1 = Based on the situation you were given, how concerned would you be about the outcome of your choice in making your selection of this product? A paired t test result showed significant differences between the high and low involvement conditions as shown in Table 3 3. The mean score for high involvement condition was 7.18 and the mean score for low involvement condition was 4.93 Pretest 3 Manipu lation 3 (Valence of O nline C onsumer R eview) P retest 3 was conducted to pre check the valence of online consumer review manipulation for the main experiment. In Pretest 3 two sets of online consumer reviews (one with 5 positive reviews and the other with 5 negative reviews ) were created respectively f or each product type (i.e., grammar software and game software) selected by Pretest 1 to be used for the main experiment I n total, four sets of 5 consumer reviews positive reviews for functional products, n egative reviews for functional
45 products, positive reviews for hedonic products, and negative reviews for hedonic products were created (refer to Appendix). The number 5 wa s chosen as it is the average number of reviews consumers generally read (Park & Le e, 2008). Each review includes line s of content with a f ont size of 10, which is typical of online consumer reviews (Park & Lee, 2008). Considering the possible interference, the quality (the number of argument s in a consumer review = 1~2) and the quantity (the number of consumer rev iews = 5; length = 3 lines) w ere controlled (Lee, Park & Han, 2007; Park, Lee & Han, 2007). To make 5 online consumer reviews for each set more natural and less artificial, actual positive and negative p roduct reviews for the product s (selected by P retest 1) on Amazon.com w ere referred and partially adopted. Amazon.com has the highest number of visitors within the consumer review/general shopping sites category (source: alexa.com). In order to pre check whether the two positive review sets and the two negative reviews sets are clearly perceived differently as intended, the study followed up with the same group of participants (n=39, 20 males and 19 females) used for Pretest 1 and 2. T he consumer reviews w ere given to the par ticipants and they were asked to evaluate the valence of each set of review s on a 9 point scale : How do you perce ive the consumer reviews above? N egative reviews P ositive reviews A paired t test w as performed to conf irm if there wa s a statistical difference in valence of reviews between the two sets. The results are shown in Table 3 4.
46 Main Experiment (Main Study I & II) The main experiment w as conducted online at Qualtrics.com. Although conducting online experiment s has some known weaknesses ( e.g., less control, distraction, difference in web browser or Internet speed by users, etc.), it clearly has advantages in convenience and multi fun ctionality. It also saves money and time. This online experiment employs a 2 x 2 x 2 design with manipulated valence (positive vs. negative), product type (hedonic vs. functional), and involvement level (high vs. low), which makes 8 unique cells with different conditions. In general, a sample size of thirty (n=30) for each cell is con sidered to be sat isfactory to conduct SEM analys e s (Ding & Harlow, 1995). In addition although there is no single rule about how large a sample needs to be for SEM analysis, in most cases, a sample size of 200+ is considered to be large enough (Kline, 200 5). Therefore, the main study aimed at collecting at least 30 college students for each of the 8 cells, which makes 240 ~250 in total (n=8x30=240) Finally, t wo hundred fifty college students participated in the online experiment The link s to access the on line experiment were sent to college students who participated voluntarily in the study through uSamp. The research panel company randomly assign ed the participants to 1 of 8 different cells ( 2 2 x 2 conditions) by sending them 8 different links based on random order. Balancing gender ratio for each cell was also considered. Finally, i n each cell, 31 or 32 participants were assigned. Conditions for each cell are shown in Table 3 5. P articipant s in each cell were instructed to go thr ough identical procedur es but in 8 different conditions. All pa rticipants w ere given a set of 5 consumer reviews and manipulated situations designed fo r each cell. T hey w ere then asked to read the
47 reviews and respond to a series of questions that measure each construct (i.e. af fective response, cognitive response, attitude toward consumer reviews, attitude toward product/brand, and purchase intention). Measurements for each construct are further discussed in the following section. Also, during the main experiment a manipulatio n check was conducted to confirm the results of P retest 1, 2, and 3 Measurement Instruments Affective Response toward Consumer Reviews Affective responses to ward consumer reviews w ere measured by AdS AM a nonverbal measurement of affective response as di scussed in Chapter 2 AdS AM was originally developed to measure affective responses to advertising stimuli ; however it has been proven to have great potential when used for various types of marketing and promotional communication (Kim & Morris, 2007; Mor ris et al 2002). Cognitive Response toward Consumer Reviews Many researchers developed measurement items to measure cognition by asking a series of bipolar scale questions (Fishbein, 1963; Kempf, 1999; MacKenzie & Lutz, 1989). For the current study, Fish value measures level brand beliefs (Bi) and attribute evaluations (Ei) w ere adopted to measure cognitive response toward consumer reviews. Cognition is generally known for its role in cr eating associations between the attitude object and attributes (Fishbein & Ajzen, 1975), so this technique was been frequently used as a practical measurement by numerous previous marketing communications studies (Kempf 1999; Kempf & Smith 1998; Marks & Kamins 1988; Smith 1993).
48 To develop the meas urement, salient attributes w ere collected for each product in the pair selected in P retest 1. Following the technique modified by K e mpf (1999) and Kim and Morris (2007 ), the m ost frequently found attributes ( as shown in Table 3 2) w ere included for the measurement for cognitive r esponse for the current study The four most salient attributes for grammar software (functional product) were accuracy, easy to use, speed, and reliability. On the other hand, the four most salient attributes for game software (hedonic product) were entertainment/fun, graphics, challenge, and stability. Attribute beliefs (Bi) w as m easured on a 9 point semantic differential scale asking the following question as sugge sted by Fishbein and Ajzen (1975) : How likely do you believe it is that the grammar software (or game software) has attribute _____? ( 1 = Z ero likelihood 9 = C ompletely certain ) For the entire study, a 9 point scale was used instead of a five or seven point scale for two reasons: ( 1) traditional five point scale is often criticized because some respondents tend to stay in the moderate options in the middle rather than choosing either a negative or positive extreme. A 9 point scale offers more sele ctions for respondents to choose from and increase s the accuracy of the measure, and (2) AdS AM the well established measure ment tool for emot ion used for the current study wa s originally designed to use a 9 point scale Because of these reasons using th e same 9 point scale for all other measures within the study may provide convenience and consistency to the current study. Attribute evaluations (Ei) w ere measured with a 9 point semantic differential scale by asking the following questi on as suggest by Fi shbein and Ajzen (1975) a nd modified by Antonides (1996) :
49 How would you evaluate the importance of attribute _____ for this grammar software (or game software) ? ( 1 = V ery unimportant 9 = V ery important ) The mean score between measured attribute le vel brand beliefs (Bi) and attribute evaluations (Ei) was calculated for each attribute and treated as items that measure the construct, cognitive response toward consumer reviews Attitude toward Consumer Reviews Attitud e toward consumer reviews w as measu red by a 3 item 9 point semantic differential scale asking the following questio n : Overall, how would you rate the consumer product reviews? ( 1 = B ad 9 = G ood / 1 = U nfavorable 9 = F avorable / 1 = D islike 9 = L ike ) Consumer Review based Attitude toward Product/Brand Attitude toward the product/brand fo rmed by consumer reviews was measured by a 3 item 9 point semantic differential scale that has frequently been used by ma ny marketing studies (Jun, Cho & Kwon, 2008; Kempf 1999; Kim & Mo rris, 2007; MacKenzie & Lutz 1989; Smith 1993) : After reading the consumer reviews, how would you rate the product/brand? ( 1 = B ad 9 = G ood / 1 = U nfavorable 9 = F avorable / 1 = D islike 9 = L ike ) Behavioral (Purchase) Intention To meas ure purchase intentions, 3 item 9 point semantic differential scale was adopted. The f ollowing question w as asked as used in many previo us marketing studies (Jun, Cho & Kwon, 2008) : If I were in the marketplace, I would buy the product. ( 1 = U nlikely 9 = L ikely / 1 = I mpossible 9 = P ossible / 1 = I mprobable 9 = P robable )
50 Analysis Strategy M ultiple R egression A nalys i s For Main Study I, m ultiple regression analys e s were conducted to examine the influences of affect and cognition on shapi ng attitudes in consumer review context. A series of multiple regression analyses was performed to compare the contribution of each indicator of independent variables (cognition and affect) on three different dependent variables consumer review attitude ( Ar ), product attitude ( Ap ), and purchase intention ( PI ) formation respectively un der eight different conditions (2 valences x 2 product types x 2 involvement levels) Considering the unique nature of each dimension of affect P, A, and D each of the m w as treated as separate independent variables for regression instead of using a combined value. Therefore, t he four independent variables that were used for the multiple regression analysis were the three dimensions of affective response P (pleasure), A (arousal), and D (dominance) and expectancy value from product SEM (Structural Equation Modeling) For Main Study II, SEM (Structural Equation Modeling) was conducted t o determine the overall re lationship among constructs SEM is a combination of multiple multivariate techniques such as multiple regression and factor analysis, and unlike most other multivariate data analysis techniques, SEM allows researchers to examine technique is especially powerful when researchers deal with multiple relationships simultaneously where one dependent variable works as an independent variable in subsequent relationships (Hair et al., 1998).
51 For SEM analysis, there are 5 logic al steps that most researchers general ly follow. (1) The researcher specifies a hypothesized model based on theory and previous research. For the current study, the hypothesized overarching model (Figure 2 4 ) that shows the overall comprehensive relationship among affective response, cognitive response, attitude toward consumer reviews, attitude toward product/brand, and purchase intention is suggested. (2) The researcher needs to determine whether the sugge sted model is identified. To be identified, the model has to be theoretically possible for the SEM program to derive a unique estimate of every model parameter. (3) The researcher needs to select the measures of the variables in the model. (4) The research er needs to conduct the analysis using a SEM program with the collected data and estimate the model. In this stage, the researcher needs to evaluate the model fit calculated by the program (further discussed below). (5) If needed, the researcher needs to s pecify the model again and evaluate the fit of the revised model with the same data set until a satisfactory model is yielded (Kline, 2005). Evaluating model fit is a critical process of the SEM analysis. In this process, researchers determine how we ll the model as a whole explains the data. Although different researchers use different set s of indices, there are set s of fit indices that are more commonly used than others in SEM analyses. (1) the model chi square (the smaller, the better; this measure is more effective for a model with small case s that have less than 200 participants .), (2) the Root Mean Square Error of Approximation (RMSEA; An RMSEA value of .05 or less is considered good, whereas 10 or higher is considered to be a bad fit.), (3) the standardized root mean square residual (SRMR; A n SRMR value of less than .08 is considered a good fit.), (4) TLI (An TLI value of higher
52 than 95 is considered a good fit.) and (5) the Bentler Bonett Index or Normed Fit Index (NFI; A n NFI value between .9 0 and .95 is considered acceptable ; a value over .95 is considered good.) (Hair et al., 1998; Kline, 2005). The final model achieved by SEM is especially effective in providing managerial as well as theoretical implications for the current study. Table 3 1. Paired t test of product type Mean SD t value d.f. Sig. Pair 1 Conference Travel Spring Break Travel 2.64 8.15 1.90 1.01 16.79** 38 .00 Pair 2 Grammar Book Cartoon Book 2.31 8.07 1.20 1.24 20.20** 38 .00 Pair 3 Grammar Software Game Softw are 2.28 8.38 1.31 .96 26.33** 38 .00 ** p < .05 (two tailed) Table 3 2 Salient attributes of grammar and game software Attributes Evaluation (%) Grammar Software Accuracy Easy to use Speed 84.61 46.15 38.46 Reliability 38.46 Game Software E ntertainment/Fun 92.30 Graphics 58.97 Challenge Stability 46.15 41.02 Table 3 3 t test of involvement level Mean SD t value d.f. Sig. High Involvement 7.18 1.09 11.21** 38 .00 Low Involvement 4.93 2.01 ** p < .05 (two tailed) Table 3 4 t test of valence of online consumer reviews Mean SD t value d.f. Sig. Negative Reviews 3.08 2.59 8.64** 38 .00 Positive Reviews 7.64 1.27 ** p < .05 (two tailed)
53 Table 3 5 Combinations of conditions for 8 cells Cell # Valence Product type Involvement level Number of participants Gender ratio (male : female) 01 P F H 31 15:16 02 P F L 31 15:16 03 P H H 31 15:16 04 P H L 31 15:16 05 N F H 32 15:17 06 N F L 31 15:16 07 N H H 31 15:16 08 N H L 32 16:16 8 250 121:129 Va lence: P (Positive) vs. N (Negative) Product type: F (Functional) vs. H (Hedonic) Involvement level: H (High) vs. L (Low)
54 CHAPTER 4 RESULTS Main Study I Manipulation Checks The 3 conditions manipulated for the current study (1) product t ype, (2) involvement level, and (3) valence were verified in the main experiment. (1) To verify the significant difference in how the participants perce ived the two different product types ( functional v s. hedonic ) manipulated by Pretest 1 the participa nts were asked the following question: How would you categorize the product in the consumer reviews above? ( 1 = primarily for functional use 9 = ) As expected, there was a significant difference detected Participants perceived the grammar check ing software as a functional product ( M ean = 3.99 SD = 2.58, t = 17.32, df = 124 ) and the game software as a hedonic product ( M ean = 6.70 SD = 2.16, t = 34.60, df = 124 ) at p < .001. (2) To verify the significant difference in how the participants perceived the two different involvement levels (high vs. low) after reading the scenario as manipulated by Pretest 2 the participants were asked the same 3 item questions asked for P retest 2: Based on the situation you were given, in selecting this product from many other choices availabl e in the market, would you say? Based on the situation you were given, how important would it be for you to make a c hoice on this product? Based on the situation you were given, how concerned would you be about the outcome of your choice in making your selection of this product?
55 A s expected the main experiment also showed that the participants perceived the two manipulated situations different ly. The mean score of the three measured items was calculated and treated as the level of involvement for each c ondition. The gap was narrower than the pretest : however the difference was still statistically significant between a high involvement situation ( M ean = 7.34 SD = 1.31, t = 62.57, df = 124), and a low involvement situation ( M ean = 6.89, SD = 1.51, t = 50 .94, df = 124) at p < .001 (3) To verify the significant difference in how the participants perceived the valence of consumer reviews (positive vs. negative), the participants were asked the same question that was asked for Pr etest 3: How do you perce ive the consumer reviews above? Negative reviews Positive reviews A s expected the main experiment also showed that the participants perceived the valence of consumer reviews differently as intended. Participants perceived the positively manipulated consumer reviews positively ( M ean = 7.80 SD = 1.55, t = 56.17, df = 123 ) and the negatively manipulated consumer reviews negatively ( M ean = 3.40 SD = 2.85, t = 13.40, df = 125 ) at p < .001. Hypothesis Testing A series of multiple regression analyses was conducted to investigate the main inquiry of the current study which is the role of affect and cognition in the impact of positive/negative online consumer reviews on brand attitude and purchase intention across 8 different conditions (2 prod uct types x 2 involvement levels x 2 valences). As discussed in Chapter 3, t he four independent variables that were used for the analysis were the three dimensions of affective response P (pleasure), A (arousal), and D (dominance) and expectancy value from product attributes ( which represents
56 cognitive response. E ach dimension of affect P, A, and D they were treated as separate independent variables for regression instead of using a combined value. A factor analy sis on PAD representing aff ective response was initially conducted to generate factor scores, and alternatively, the scores were used as independent variables as surrogates for the raw data. Although there ha ve been some pros and cons regarding this method (Horst, 1965; Kerlinger & Pedhazur, 1973; Kukuk & Baty, 1979; Rummel, 1970), using factor scores is considered to be theoretical ly superior in term s of ease of interpretation and generalizability Using factor scores is also considered to increase the research sensitivity (Cohen & Cohen, 1983). Three sets of separate multiple regressions were conducted to examine the influence of the four independent variables on the three dependent variables consumer review attitude ( Ar ) f ormation p roduct a ttitude ( Ap ) f ormation and p urchase i ntention ( PI ) f ormation respectively. RQ1, RQ4. Consumer Review Attitude ( A r ) Formation (H 5 ~ H8 ) Hypothesis 5 (Functional product + High involvement) H 5 was supported only in positive review condition. As shown in Table 4 1, cognitive response was a dominant influencer on consumer review attitude when subjects were exposed to positive online consumer reviews (coefficient = .81, p < .01). This finding is consistent with Kempf (1999), Batra and Stephens (1994), and Kim and Morris (2007). The hypothesis however, was not supported in negative review condition. Two of the dimensions of affect (P: coefficient = .66, p <.01, and A: coefficient = .38, p < .01) were the only predictors of consumer review attitude formation while the influence of cognition was not significant.
57 Hypothesis 6 (Functional product + Low involvement) As expected, the effect of cognition ( coefficient = .50, p < .01) on consumer review attitude ( A r ) seems to be greater and more significant than affect (P: coefficient = .39, p < .05, bu t A & D: insignificant) when subjects were exposed to positive reviews: however, affect (P) was the only significant predictor of consumer review attitude ( A r ) formation when the participants were exposed to negative reviews ( coefficient = .58, p < .05). H ypothesis 7 (Hedonic product + High involvement) In the case of positive review condition, D of affect seems to exert a certain level of influence ( coefficient = .30, p < .05) on c onsumer r eview attitude ( A r ) formation: however, the results again found mo re significant influence from cognitive response ( coefficient = .57, p < .01). In the case of negative review condition, none of the affective dimensions were found to be significant predictors of c onsumer r eview attitude ( A r ) formation while cognitive res ponse showed a certain level of influence ( coefficient = .49, p < .05) on the dependent variable. Therefore, H 7 was not supported. Hypothesis 8 (Hedonic product + Low involvement) Again, cognitive response showed a more significant influence ( coefficient = .71, p < .01) on c onsumer r eview attitude ( A r ) when respondents were exposed to positive reviews, which may conflict with findings from Kempf (1999), Batra and Stephens (1994): however, as expected, in the case of negative consumer review exposure, affe ct (P: coefficient = .38, p < .1) showed a certain level of significant influence whereas the influence from cognitive response was not significant. Overall, as shown in Table 4 1, hypotheses were partially supported depending on the valence of consumer re views. In other words, the influence of cognitive and affective
58 responses on consumer review attitude ( Ar ) was found to be moderated only by the valence of consumer reviews. However, the moderating effects of the other two variables product type and invo lvement level were merely detected. RQ2, RQ4. Product Attitude ( A p ) Formation (H 9 ~ H1 2 ) Hypothesis 9 (Functional product + High involvement) As shown in Table 4 2 significant impacts from both affective (P: coefficient = .38, p < .01) and cognitive r esponses ( coefficient = .62, p < .01) on p roduct a ttitude ( A p ) f ormation were detected when participants were exposed to positive consumer reviews: however, the effect of cognition seemed to be greater than that of affect. The difference in coefficient siz e between P and cognition was .24. On the other hand, when exposed to negative reviews, participants affective responses (P: coefficient = .74, p < .01, and A: coefficient = .33, p < .01) were a more powerful predictor for product attitude ( A p ) formation than cognitive response ( coefficient = .22, p < .05). Therefore, the results support H 9 and coincide with previous findings by Kempf (1999) and Batra and Stephens (1994) under positive review condition only. Hypothesis 10 (Functional product + Low involv ement) Results were similar to those found for H 9 When respondents were exposed to positive reviews, as expected, cognitive response ( coefficient = .55, p < .01) was a dominant influencer on p roduct a ttitude ( A p ) f ormation : however, for negative review co ndition, it was found that affect (P: coefficient = .79, p < .01) was the dominant and the only variable that influenced p roduct a ttitude ( A p ) f ormation The finding is consistent with s (1994) study, but does not support H 10 for the cur rent study in negative review condition.
59 Hypothesis 11 (Hedonic product + High involvement) For positive review condition, influences from both affect and cognition w ere detected. Interestingly, the only affective dimension that was significant was D ( coef ficient = .36, p < .05): however, in terms of the size of the coefficient, cognition ( coefficient = .60, p < .01) was a more powerful influencer than D of affect. Consequently, the results do not support H11: however, for negative review condition, the res ults were the opposite. Influences from both affect (P: coefficient = .52, p < .01) and cognition ( coefficient = .38, p < .01) were detected: however, the size of the effect of affect (P) appears to be greater than cognition in forming p roduct a ttitude ( A p ) The difference was .14. Hypothesis 1 2 (Hedonic product + Low involvement) In positive review condition, although some degree of significant influence from affect (P: coefficient = .31, p < .05, and D: coefficient = .28, p < .05) was detected, cognition ( coefficient = .53, p < .01) seemed to exert more influence in forming p roduct a ttitude ( A p ) Therefore, the results under positive review condition do not support H1 2 In negative review condition, however, affect (P: coefficient = .56, p < .01) was the o nly significant influencer of p roduct a ttitude ( A p ) formation, which supports H 12 Overall, the results of multiple regression analysis on the influence of affect and cognition on product attitude ( Ap ) formation showed very similar pattern found in consume r review attitude ( Ar ) formation. As shown in Table 4 2 hypotheses were again partially supported depending on the valence of consumer reviews, and the influence s from the other moderating variables product type and involvement level were merely detec ted.
60 RQ3, RQ4. Purchase Intention ( PI ) Formation (H1 3 ~ H1 6 ) Hypothesis 1 3 (Functional product + High involvement) As shown in Table 4 3 when participants were exposed to positive consumer review condition, significant impacts from cognitive responses ( coefficient = .49, p < .01) on p urchase i ntention ( PI ) f ormation w ere detected and were greater than the impact from affective responses (P: coefficient = .29, p < .05). Therefore, H1 3 was supported under positive review condition: h owever, when participa nts were exposed to negative review condition the effects of both affective response (P : coefficient = .69, p < .01, and A : coefficient = .33, p < .01 ) and cognitive response ( coefficient = .23, p < .05) were detected to some degree However P was found to be the most powerful predictor of all independent variables (at p < .01 ) Therefore, the results for negative review condition do not support H 1 3 Hypothesis 1 4 (Functional product + Low involvement) When participants were exposed to positive reviews, cognition ( coefficient = .54, p < .01) seemed to be the only independent variable that s ignificant ly influenced the p urchase i ntention ( PI ) f ormation On the other hand, under negative review condition, P was found to be the most powerful predictor of p urc hase i ntention ( PI ) that was significant ( coefficient = .79, p < .01 ). Therefore, again, H1 4 was supported only when the participants were exposed to positive reviews. Hypothesis 1 5 (Hedonic product + High involvement) For positive review condition, cogni tion ( coefficient = .70, p < .01) was the only predictor that significantly influenced the p urchase i ntention ( PI ) f ormation which conflicted with H15. For nega tive review condition, although influences from both affect (P: coefficient = .53, p < .01, and D: coefficient = .26, p < .05) and cognition ( coefficient
61 = .38, p < .01) were detected P of affect was found to be the most dominant influencer. Therefore, H1 5 was supported in negative review condition only. Hypothesis 1 6 (Hedonic product + Low involv ement) In positive review condition, the effects of affective (P) and cognitive responses were fairly balanced (coefficient = .45 p < .01 ) T he impact from cognitive response was slightly larger than P of affect, but the difference was minimal (.003) On the other hand, when participants were exposed to negative reviews, affective response (P: coefficient = .50, p < .01, D: coefficient = .27, p < .10) was found to be a significant influencer of p urchase i ntention ( PI ) formation. Therefore, H1 6 was supporte d only under negative review condition. Again, the results of multiple regression analysis on the influence of affect and cognition in predicting purchase intention ( PI ) showed the same overall pattern found in consumer review attitude ( Ar ) and product att itude ( Ap ) formation. As shown in Table 4 3 hypotheses were again partially supported depending on the valence of consumer reviews When participants were exposed to positive consumer review s, cognition shadowed the influence of affect i n predicting purch ase intention (PI) regardless of product type and involvement level. However when people were exposed to negative consumer review s affect dominated over cognition for predicting purchase intention (PI) regardless of product type and involvement level. In any cases, P (pleasure) was always the most dominant influencer among the three dimensions of affect in predicting purchase intention fo rmation
62 Main Study II Overall Relationship among Constructs CFA (Confirmatory Factor Analysis) and SEM (Structural E quation Modeling) analysis by AMOS (19 th edition) were conducted to test the h ypothesized model suggested in C hapter 2 and to examine t he overall relationship among the 5 constructs (latent variables: a ffective r esponse toward c onsumer r eviews c ognitive r esponse toward c onsumer r eview s, a ttitude toward c onsumer r eviews [ Ar ], c onsumer r eview based a ttitude toward p roduct/ b rand [ Ap ], and b ehavioral ( p urchase) i ntention [ PI ]). The SEM analysis of the current study followed the rest of the SEM procedures state d in C hapter 3 To test the hypothesized model, a ll data from the 8 cells w ere combined and analyzed. The analysis was conducted using the following procedures : (1) the reliability and validity of each construct was examined, (2) the model fit calculated b y AMOS was evaluated, and, ( 3) based on examination of the measurement and structural parameters the hypothesized model and alternative models were compared to determine the model that the experiment data support ed the most (Kline, 2005). All Skewness an d Kurtosis values for each item were within a range of 1.96 All Skewness values were between 72 and 02 and all Kurtosis values were between 1. 55 and 06, which indicate s the normality of the data. The descriptive results of each measurement item an d correlation matrix of measurement items are shown in Table 4 4 and Table 4 5 Table 4 4 especially shows the difference in the mean scores of each measured item based on the valence of consumer reviews (positive vs. negative). T he results of validity a nd reliability check s are discussed below.
63 Reliability and validity Reliability and validity of the construct measures were assessed using the combined data from all 8 cells. The reliability was assessed based on Cronbach s alpha which indica tes the in ternal consistency among items by measuring each construct. Results confirmed that all construct measures were reliable producing Cronbach s alpha of .90 for cognitive response, .98 for attitude toward consumer review, .99 for attitude toward product, and .98 for purchase intention, which are above the minimal standard of 8 0 suggested by Nunnally (1978). Discriminant and convergent validity two of the construct validities w ere also used to assess the construct measures. Discriminant validity refers to the extent to which a measure differentiates a particular concept from other concepts (Singleton & Straits, 1999). Discriminant validity of the construct measures was confirmed by testi ng the pairwise correlations in Table 4 6 As shown i n the table, the corr elations between constructs ranged from 53 to 84, which confirm s discriminant validity (Affect Cognition = .57, Affect Ar = .53, Affect Ap = .67, Affect PI = .66, Cognition Ar = .70, Cognition Ap = .77, Cognition PI = .78, Ar Ap = .84, Ar PI = .82). Correlation of less than .85 is a generally accepted criterion for decent discriminant validity. Although the correlation between the two constructs, Ap and PI is evidently higher than the criterion ( Ap PI = .95) they both have been strong ly established by numerous previous studies, and they do not serve as independent variables in the current study, therefore, there should be no multicollinearity issue for the current study caused by this high correlation. Multicollinearity issue generally occurs when there are high c orrelations among the independent variables (Hair et al., 1998 ).
64 Convergent validity refers to the correspondence of results when a construct is measured in different ways (Singleton & Straits, 1999) and it was assessed by che cking whether all items factor loadings on their corresponding constructs were significant or not ( Anderson & Gerbing 1988 ). The results, shown in Table 4 7 indicate that all items significantly loaded to the corresponding constructs at a .0 0 1 level : th erefore, adequate convergent validit y was achieved Confirmatory F a ctor Analysis (CFA) A confirmatory factor analysis was conducted with all combined data across the 8 different cells to check the measurement model. The g oodness of fit indices of the init ial model ( 2 66.74 df = 94 GFI = .89, TLI = 97 SRMR =. 03 RMSEA = .0 9 ) generated a marginally acceptable overall fit but indicated needs to respecify the initial measurement model for improvement. Based on the information provided by the modification indice s, the covariance relationships between variables in the initial model were altered to improve the model fit. The alterations were made only when the changes were theoretically justifiable because theory based respecification can reduces the possibility of sampling error to improve the goodness of fit (Anderson & Gerbing 1988). The first noticeably high modification index was between the item #3 (Ar3: likability of consumer reviews) of consumer review attitude ( Ar ) and the item # 2 ( Ap2 : favorability of pro duct ) of p roduct attitude ( Ap ). Numerous researchers found the theoretical relationship between attitude formation toward a product/brand message and attitude toward product/ brand (Jun, Cho, & Kwon, 2008; MacKenzie, Lutz, & Belch, 1986; Perloff, 2008). The refore, the error covariance for the two items was freed to estimate.
65 The next high modification index reported was between the item #3 (Ar3: likability of consumer reviews) of consumer review attitude ( Ar ) and the item # 2 ( PI2 : possibility of purchase ) of consumer review based purchase intention ( PI ). Likewise, error covariance of those two items w as freed for the same theoretical background as the first respecification. After those two respecifications, there were considera ble increases in the model fit s especially the chi square and RMSEA ( 204.77 df = 91 GFI = .91, TLI = 98 SRMR =. 03 RMSEA = .0 7 ) Figure 4 1 shows the final measurement model with factor loadings for each item and the construct correlations. Structural Equ ation Modeling (SEM): H1, H2, H3 H 4 SEM analysis was conduc ted to determine the overall relationships among the constructs. SEM is a combination of multiple multivariate techniques such as regression et al., 1998: p.578). F or SEM analysis, t he current study especially adopted a MIMIC model as an alternative to multiple sample analysis ( Kline, 2005 ; Reisenzein 1986 ) on the bas is of the findings from Main Study I The Main S t udy I tested the moderating effects of three variables product type, involvement level, and valence of consumer reviews. However, the series of multiple regression analysis found a dominant moderating effect of valence of consumer reviews, whereas the effects from the other two mo derating variables were merely detected. The refore, Main Study II adopted a MIMIC model to estimate group differences (valence of consumer reviews: positive vs. negative) on latent variables. MIMIC model is especially useful when factors with effect indic ators are regressed on
66 one or more dichotomous cause indicators that represent group membership, such as one coded 0 = treatment and 1 = control (Kline, 2005: p.307). The main experiment of the current study manipulated valence of consumer revi e ws as c on sumer review s generally contain either positive or negative information about products/ brands/ services in its nature (Haywood, 1989). T he valence of the information may exert critical role in consumers and p resumably generate critical difference in results of the current study When there are two sets of data due to experimentally manipulated dichotomous situations, the MIMIC model technique allows researchers to conduct the SEM analysis with combined data set instead of dividing the da ta for two separate analyses ( in case of the current study, positive review group: n = 124, negative review group: n = 126, number of combined data: n = 250) The technique is useful in examining the difference s between two data sets as it uses experimenta lly manipulated dummy variable that represents the comparison of two dichotomous situations, which is valence of consumer reviews (positive vs. negative) in case of the current study Experimentally manipulated situations ( i.e. positive vs. negative review s ) were treated as a covariate in the SEM model tested for the current study The situations were named as a new variable Valence single indicator dichotomy (negative review situation coded as 0 and positive review situation as 1 ) and used as a n o bserved, exogenous variable. The initial model in Figure 4 2 hypothesize s the relationship among all constructs including affective response, cognitive response, attitude toward consumer review ( Ar ) attitude toward product ( Ap ) and purchase intention ( PI ) T he variable named group is a covariate. The initial model include s six paths ( valence affect valence cognition,
67 affect attitude toward consumer review cognit ion attitude toward consumer review, attitude toward consumer review attitude towa rd product attitude toward product purchase intention ). Since a MIMIC model is a covariate structure, the path coefficient s for the direct effects of Valence provide information about the degree to which the difference between positive and negative re view predicts the factors (Kline, 2005). As indicated in Table 4 9 a ll paths showed significant path coefficients (at .a 05 level) ranging from .40 to .96 (valence affect : .66, valence cognition: .51, affect consumer review attitude: .59, cognit ion consumer review attitude: .40, consumer review attitude product attitude: .85, product attitude purchase intention: .96) however the model fits were not acceptable as indicated in Table 4 8 481.26 TLI = .93, SRMR = 15 RMSEA = .1 2, NFI = .93 ) Especially, SRMR and RMSEA indices were poor. Examination of the modification index suggested that a new path (cognition affect in dashed arrow) be added to the model as shown in the Mod1 in Figure 4 2 for respecification. There was a huge improvement in and SRMR indices, but still unsatisfactory 379.29 TLI = .95, SRMR = 09 RMSEA = .1 0, NFI = .94 ). As the new path was added (cognition affect), the coefficient of the existing path (cognition Ar ) dropped dramatically from .40 to .04 and becam e insignificant as shown in Table 4 9 Finally, t he Amos modification index suggested to add another new path (cognition Ap in dashed arrow ) t o the Mod1 as shown in Mod2 in Figure 4 2 Because the suggestion was theoretically meaningful, the Mod1 was rev ised allowing for this new loading showing significant increase in and SRMR indices and acceptable model fits 312.48 TLI = .96, SRMR = 05 RMSEA = 09, NFI = .95 ) as indicated in Table 4
68 8 As additional change in the model did not increase the fit any further, no more respecification was made for parsimoniousn ess of the model. As shown in Figure 4 3 the final model indicates that affective response dominantly influences attitude toward consumer review (s tandardized path coefficients = .8 7 ) while cognitive response has no significant influence on attitude towar d consumer review ( .02 ) except some indirect effect through affective response only. Table 4 1 Consumer Review Attitude ( Ar ) f ormation Conditions DV IV ( ) R F P A D Cog.( ) P FH A r .21* .19* .06 .81*** .67, 16.10*** P FL A r .39** .00 .06 .50*** .64, 11.64*** P HH A r .21 .04 .30** .57*** .54, 7.48*** P HL A r .25* .05 .02 .71*** .82, 29.91*** NFH A r .66* ** .38*** .05 .16 .72, 17.64*** NFL A r .58** .10 .03 .19 .55, 7.87*** NHH A r .04 .10 .23 .49** .27, 2.36*** NHL A r .38* .01 .14 .29 .40, 4.46*** P FH = Positive review + Functional product + High involvem ent P FL = Positive review + Functional product + Low involvement P HH = Positive review + Hedonic product + High involvement P HL = Positive review + Hedonic product + Low involvement N FH = Negative review + Functional product + High involvement N FL = Negati ve review + Functional product + Low involvement N HH = Negative review + Hedonic product + High involvement N HL = Negative review + Hedonic product + Low involvement p < .1 0 ** p < .05, *** p < .01 = the most dominant predictor
69 T able 4 2 Pr oduct Attitude ( A p ) formation Conditions DV IV ( ) R F P A D Cog.( ) P FH A p .38*** .13 .03 .62*** .62, 10.49*** P FL A p .27 .09 .07 .55*** .57, 8.60*** P HH A p .17 .06 .36** .60*** .58, 8.94*** P HL A p .31** .08 .28** .53 *** .79, 24.29*** N FH A p .74*** .33*** .04 .22** .87, 44.91*** N FL A p .79*** .08 .15* .13 .83, 31.19*** N HH A p .52*** .06 .20* .38*** .68, 14.09*** N HL A p .56*** .08 .28** .23 .65, 12.59*** PFH = Positive revi ew + Functional product + High involvement PFL = Positive review + Functional product + Low involvement PHH = Positive review + Hedonic product + High involvement PHL = Positive review + Hedonic product + Low involvement NFH = Negative review + Functional product + High involvement NFL = Negative review + Functional product + Low involvement NHH = Negative review + Hedonic product + High involvement NHL = Negative review + Hedonic product + Low involvement p < .1 0 ** p < .05, *** p < .01 = the most dominant predictor Table 4 3 Purchase Intention ( PI ) formation Conditions DV IV ( ) R F P A D Cog.( ) P FH PI .29** .06 .25 .49*** .57, 8.58*** P FL PI .23 .02 .08 .54*** .51, 6.74*** P HH PI .04 .04 .16 .70*** .53, 7.33*** P HL PI .45*** .13 .12 .45*** .78, 23.48*** N FH PI .69*** .33*** .02 .23** .82, 30.92*** N FL PI .79*** .05 .14* .14 .83, 31.74*** N HH PI .53*** .09 .26** .38*** .74, 18.59*** N HL PI .50*** .05 .27* .27 .61, 10.47*** PFH = Positive review + Functional prod uct + High involvement PFL = Positive review + Functional product + Low involvement PHH = Positive review + Hedonic product + High involvement PHL = Positive review + Hedonic product + Low involvement NFH = Negative review + Functional product + High invol vement NFL = Negative review + Functional product + Low involvement NHH = Negative review + Hedonic product + High involvement NHL = Negative review + Hedonic product + Low involvement p < .1 0 ** p < .05, *** p < .01 = the most dominant predictor
70 Table 4 4 Descriptive statistics of measurement items Mean (Positive, n=124) Mean (Negative, n=126) Mean (Combined, n=250) SD (Combined, n=250) P 7.00 3.83 5.40 2.83 A 4.99 4.69 4.84 2.57 D 5.73 4.93 5.32 2.40 Cog1 7.42 5.67 6.54 1.85 Cog2 6.9 8 5.48 6.22 1.84 Cog3 7.15 5.43 6.29 1.89 Cog4 7.55 5.42 6.48 1.96 Ar1 7.85 4.56 6.19 2.94 Ar2 7.84 4.32 6.06 2.99 Ar3 7.70 4.33 6.00 3.01 Ap1 7.78 3.76 5.76 3.11 Ap2 7.72 3.64 5.66 3.09 Ap3 7.69 3.71 5.69 3.11 PI1 7.31 3.53 5.40 3.19 P I2 7.57 3.70 5.62 3.05 PI3 7.34 3.45 5.38 3.13 P: Pleasure A: Arousal D: Dominance Cog: Cognitive Response Ar : Attitude toward Consumer Reviews Ap : Attitude toward Product PI : Purchase Intention Table 4 5 Correlation matrix of measurement items P A D Cog1 Cog2 Cog3 Cog4 P 1 .00 A .27** 1 .00 D .28** .10 1 .00 Cog1 .6 8 ** .12 .28** 1 .00 Cog2 .6 2 ** .17** .28** .84** 1 .00 Cog3 .65** .18** .3 4 ** .8 9 ** .8 7 ** 1 .00 Cog4 .67** .12 .29** .88** .8 4** .88** 1 .00 Ar1 .69** .1 4 .1 9 ** .7 1 ** .6 3 ** .6 6 ** .67** Ar2 .69** .13* .19** .65** .6 0 ** .6 2 ** .65** Ar3 .70** .1 4 .20** .70** .61** .6 5 ** .67** Ap1 .8 4 ** .18** .3 2 ** .73** .67** .72** .76* Ap2 .8 4 ** .18** .32** .73** .6 9 ** .7 2 ** .7 6 ** Ap3 .8 1 ** .19** .33** .74** .6 9 ** .72** .74** PI1 80 ** .19** .31** .75** .68** .7 3 ** .74** PI2 .80** .18** .3 3 ** .7 3 ** .67** .72** .75** PI3 .8 1 ** .2 1 ** .3 2 ** .74** .69** .7 5 ** .7 8 **
71 Table 4 5 Continued Ar1 Ar2 Ar3 Ap1 Ap2 Ap3 PI1 PI2 PI3 P A D Cog1 Cog2 Cog3 Cog4 Ar1 1 .00 Ar2 .9 3 ** 1 .00 Ar3 .9 3 ** .95** 1 .00 Ap1 80 ** .80** .81** 1 .00 Ap2 .81** .8 2 ** .81** .96** 1 .00 Ap3 .7 9 ** .79** .82** .9 6 ** .96** 1 .00 PI1 .78** .7 7 ** .80** .9 2 ** .9 2 ** .93** 1 .00 PI2 .79** .7 8 ** .78** .9 3 ** .93 ** .9 3 ** .9 4 ** 1 .00 PI3 .78** .7 8 ** 80 ** .91** .9 3 ** .93** .9 6 ** .95** 1 .00 P: Pleasure A: Arousal D: Dominance Cog: Cognitive Response Ar : Attitude toward Consumer Reviews Ap : Attitude toward Product PI : Purchase Intention Correlation is signi ficant at the .05 level (2 tailed). ** Correlation is significant at the .01 level (2 tailed). Table 4 6 Summary s tatistics and c orrelation among constructs Affect Cognition Ar Ap PI Mean SD Affect 1 .00 5.1 9 1.8 1 Cognition 5 7* 1 .00 6.38 1.79 Ar 53 7 0* 1 .00 6.0 9 2.91 Ap 6 7* 7 7* 84 1 .00 5.70 3.0 6 PI 6 6* 7 8* 82 9 5* 1 .00 5.4 7 3.07 *** Correlation is significant at the .01 level (2 tailed).
72 Table 4 7 Results of CFA (Group number 1 Default model) Factor loading Standard Error Critical Ratio (t value) P Affect .93 A Affect .2 6 .06 3.93*** D Affect 3 3 .06 5.20*** Cog1 Cognition .94 Cog2 Cognition 90 04 25.81*** Cog3 Cognition .94 03 30.35*** Cog4 Cognitio n .93 04 29.09*** Ar1 Ar .95 Ar2 Ar .97 03 38.48*** Ar3 Ar .98 03 40.86*** Ap1 Ap .97 Ap2 Ap .98 02 52.01*** Ap3 Ap .98 02 50.22*** PI1 PI .98 PI2 PI .97 02 43.89*** PI3 PI 98 02 50.96*** Goodness of fit statistics: 171.56, TLI = .98, GFI = .93, RMSEA = .06
73 Figure 4 1 Final measurement model
74 Model Tested Model Initial Mod1 Mod2 (final) Figure 4 2 Path diagrams of competing model s Table 4 8 Fit Indices of competing models Model d.f. TLI SRMR RMSEA NFI Initial 481.26 107 .93 .15 .12 .93 Mod1 379.29 106 .95 .09 .10 94 Mod2 (final) 312.48 105 .96 .05 .09 .95 Criterion > .90 < .05 < .08 > .90
75 Table 4 9 St andardized path coefficients of competing models Exogenous Endogenous Initial Mod1 Mod2 (final) Valence Affect .66** .34** .32** Valence Cognition .51** .51** .52** Affect Ar .59** .83** .87** Cognition Ar .40** .04 .02 Ar Ap .85** .86** .57** Ap PI .96** .97** .97** Cog nition Affect N/A .66** .68** Cog Ap N/A N/A .39** ** Significant at .05 level Goodness of fit statistics: 312 48, d.f. = 105, TLI = .96, SRMR = .05, RMSEA = .09, NFI = .95 ** Significant at .05 level Figure 4 3 Final model with standardized path coefficients
76 CHAPTER 5 SUMMARY AND DISCUSSI ONS Summary Along with the prosperity of the Internet, W OM (or eWOM) has become one of the most powerful forces emerg ing in marketing today as it is widely accepted by consumers as a critical information source. As eWOM attracts more attention from people marketers are also attempting to better understand the new phenomenon that could have critical impact on their brand s and compan ies In spite of some limitations to be discussed in C hapter 6 the current study made some contribution s to advance the understanding o n the process and the influence of eWOM (more concretely, on line consumer reviews as a most common form of eWOM) o n consumer s attitude and purchase intention formation. More specifically, t h e current study was conducted to compare the contribution of affect (P, A, and D) and cognition on the predict ion of consumer review attitude ( Ar ), product attitude ( Ap ), and purchase intention ( PI ) Until scholars started to realize the important role of affect in forming attitude in the 1980s, t he role of affect had long been ignored and underestimated by schola rs and r esearch had been centered mainly around people s cognitive proc essing, which was considered a dominant force over affect in forming people s attitudes (Lazarus, 1982; Lazarus, 1984; Tsal, 1985). The findings of the current study are mainly two fol d. First, a series of multiple regression analyses was conducted to investigate the main inquiry of the current study which is th e role of affect and cognition o n the impact that online consumer reviews have on brand attitude and purchase intention acros s various conditions (2 product types x 2 involvement levels x 2 valences). I n conclusion
77 the regression analyses detected very powerful moderating effects of the valence of consumer reviews on the role of affect and cognition. Researchers have investiga ted the different roles of affect and cognition in the evaluation process across different conditions in various marketing contexts. Most of the previous studies tested the moderating effect of product type and involvement level in the context of product t rial, corporate identity, advertisements, and the like Kempf (1999) detected the moderating effect of product type as he found affect to be a dominant influencer in hedonic product situation s whereas cognition was a dominant influencer in functional produ ct situation s On the other hand, Batra and Stephens (1994) found a moderating effect of involvement situation s that showed the role of affect may be more influential in shaping brand attitudes in low involvement situation s whereas cognition may be a more powerful predictor in high involvement situation s The current study added and examine d an additional moderating variable to this rather inconsistent line of research findings : the valence of online consumer reviews ( i.e. positive e WOM vs. negative e WOM) Unlike other marketing communications, WOM may contain negative information about a product or brand (Haywood, 1989) so the valence of information may be assumed to be critical in influencing how cognition and affect interplay to shap e attitudes and pur chase intention s As noted earlier and as expected the current study clearly found a consistent overarching pattern of effects of co gnition and affect on attitude / purchase intention formation. The series of the multiple regression analysis found a domi nant moderating effect from valence of consumer review whereas the effects from the other moderating variables product types and involvement levels were merely detected. In positive
78 consumer review situation, cognition dominated affect for predicting c onsumer review attitude ( Ar ), product attitude ( Ap ), and purchase intention ( PI ) regardless of product type and involvement level. On the other hand, when people were exposed to negative consumer review situation, the results were totally the opposite, i.e ., affect, in turn, dominated over cognition for predicting consumer review attitude ( Ar ), product attitude ( Ap ), and purchase intention ( PI ) regardless of product type and involvement level. Among the 3 dimension s of affect ( P, A, and D ) P (pleasure) dom inated as the most important influencer in predicting attitude and purchase intention formation Second, based on the interesting findings from the multiple regression analysis, the current study also conducted SEM analysis to examine the overall relatio nship among the five constructs (affective response toward consumer reviews, cognitive response toward consumer reviews, consumer reviews attitude [ Ar ], product/brand attitude [ Ap ], and behavioral (purchase) intention [ PI ]) The analysis found the predomi nant direct influence of affect on consumer review attitude (coefficient = .84 at p < .05) compared to cognition (coefficient = .01 at p > .05). It also detected the role of affective response as a mediator for cognitive response. When a new path from cogn ition to affect (Cog Affect, coefficient = .66 at p < .05) was added for model respecification, the path coefficient (direct effect) from cognition to consumer review attitude (Cog Ar ) almost completely disappeared (.40 at p < .05 .04 at p > .05). In other words, the direct effect of cognition on consumer review attitude became insignificant and cognition was only mediated by affect. The finding is nearly opposite to those of Lazarus ( 1984 ) and Tsal ( 1985 ) who maintain affective re sponse s are always mediated by cognition
79 The results of the SEM analysis, however suggest that the effect of cognition should not be neglected either. In the final respecified model of the SEM analysis, cognition is still a powerful influencer of affect (coefficient = .67 at p < .05) as well as important predictor of product attitude ( Ap ) (coefficient = .36 at p < .05) which advertisers would be more interested in because of its high correlation with purchase intention ( PI ). In addition, standardized indirect effect of cogn itive response on purchase intention was .62 (at p < .05), which is fairly high. Taken together, it is assumed that in terms of consumer review attitude ( Ar ), affect is a more direct and dominant predictor, whereas cognition seems to play a more critical and immediate role in forming product attitude ( Ap ) and purchase intention ( PI ). Think Positive, Feel N egative! Th e current study provides various theoretical and managerial implications that should be carefully considered by practitioners and researchers First, t h e findings of the study imply that advertis ers need to take more care of, and pay close r attention to eWOM, especially for negative consumer reviews since it is assumed that consumers may respond more emotional ly when they are exposed to negati ve reviews rather than positive reviews. It may be an important finding because, according to prior research, negative r eviews with more emotional opinion s can be more powerful and effective than logical positive opinions (Yang & Fang, 2004) Also, accord ing to a recent survey ( Cone Communications, 2011), eighty percent of consumers have changed their purchase decision s based solely on a negative consumer review found online. No advertiser can afford to ignore negative consumer reviews about their product s in today s interconnected world
80 While good c onsu mer services ha ve always been necessary for successful marketing, th e finding s of the current study provide yet another reminder of the importance of good communication skills used by marketers with their consumers in today s marketplace which is increasingly dominated by eWOM ( Kim, Lee, & Ragas 2011) By resp on ding more swiftly to negative online consumer reviews, marketers may be able to limit the amount of negative reviews and ultimately turn a negativ e into a positive ( Kim, Lee, & Ragas 2011) Second, another managerial implication is that advertisers need to understand the overall mechanism of how consumers form their attitude and purchase intention when responding to consumer reviews in differen t si tuations. According to the current study, affect may play a more significant role in forming attitude toward consumer reviews in negative condition s whereas consumers t end to use more cognition when exposed to positive reviews regardless of product type an d involvement level The results seem to be very consistent with the findings of Kim, Lee, and Ragas (2011) who content analyzed 840 actual consumer reviews from the consumer review writers point of view. The researchers found that when consumers are nega tive about a product/brand, their reviews tend to be less logical and more emotional than when they are writing positive review s The current study found a consistent pattern from the consumer review readers point of view a s well T h ese findings should be carefully considered and used strategically by advertisers to more appropriately respond to consumer s feedback or reviews. I t would be very important for advertisers to understand the status of consumers before they create either cognition or emotion bas ed communication strategies to deal with
81 consumers or the consumer reviews. When advertisers respond to negative reviews, they should understand that the consumers involved ( whether they are the readers or the writers of consumer reviews) may be highly emo tional A l ogical approach based on cognition based strategy may be more effective in dealing with the situation. On the other hand, when exposed to positive reviews, the current study suggests that consumers tend to use more cognition than affect when f orming their attitudes. W hile providing additional cognition based information on products, advertisers should consider strategies used to direct the consumers cognition based consumer review attitude to a more overall favorable attitude toward the compan y s products/brands. Third, although the importance of the role of affect and cognition in forming attitudes varies based on the valence of consumer reviews, advertisers need to understand th e balanced roles of affective and cognitive response s to faci litate the formation of a good product attitude ( Ap ). As detected by the SEM analysis in general situation s an affective response should be highlighted more for its dominant direct influence on consumer review attitude ( Ar ) ; however, when it comes to the influence on product attitude ( Ap ) formation consumers may be more influenced by cognitive processing. B ecause online consumer reviews are direct information from other consumers regarding a product /brand, and consumers usually refer to the online consum er review information when they are close to making a purchase, it may be assumed that consumers tend to make more cognition based decisions to avoid risk s Also, t he indirect effect of cognition on consumer review attitude ( Ar ) should not be neglected or underestimated
82 Fourth, the current study found that affect may play a more critical role in forming attitudes when consumers are exposed to negative information (consumer reviews) where as cognition is assumed to be more influential when exposed to positiv e information. The findings may be t heoretical ly linked to the theory suggested by psychologists, Fiske and Taylor (1984) The theory suggests that when evaluating new information and making decisions, people often tend to rely on more ti me efficient effort efficient strategi es for information processing to save their cognitive energy. Therefore, people often use their existing knowledge structure rather than rationally thinking and making cognition based decision s when it is not required (Fi sk e & Taylor, 1984). eWOM or online consumer reviews are informatio n that people actively seek to ultimately make better informed decisions. Based on this, it may be assumed that people use cognition based rational thinking only when they see positive information about t he product they are interest in ; h owever, when people are exposed to negative reviews which may be a sign of not needing further careful elaboration of information, it is assumed that people tend to save their cognitive energy and becom e emotional. According to ELM (Elaboration Likelihood Model) by Petty and Cacioppo ( 1981 ) people tend to take either central route or peripheral route to persuasion based on their level of motivation. Less motivated by negative consumer reviews, people ma y take the peripheral route with no serious message elaboration or thinking too much about message arguments and process messages based on simple affective cues at a very heuristic level
83 T here should be further investigation on the linkage and how and w hy affect and cognition influence people s information processing and attitude forming differently in different valence situation s
84 CHAPTER 6 LIMITATIONS AND SUGG ESTIONS While th e current study makes several unique contribution s to understanding t he relationship between the role of cognition/affect and attitude formation, a s with any study, there are several limitations that need to be noted. First, in this study, only two products (Grammar checking vs. Game software) were chose n to represent functional and hedonic product s, respectively There could be some controversy, however, on the ability of the two products to represent the two product type s and to generalize the results. Therefore, replications are needed through future r esearch to enhance validity. Second, the current study only adopted a few variables For future research, adding additional variables should be considered to examine any significant relationship among cognition/affect, online consumer reviews and other va riables. Over the past few years, researchers have begun to explore eWOM in the context of new online social media ( e.g., Facebook, T witter, etc. ) Also, there are many other variables that were not tested by the current study, but may be theoretically rel ated to the important constructs used in the study ( e.g., perceived risk, intention to seek out WOM, source credibility etc. ). Further analyses using various research variable s and methods to replicat e and extend the different dimension s of the topic of t he current study would help compliment this growing line of inquiry Third, the current study used college student s for the sample. Although this homogeneous sampling was considered effective in controlling other extraneous variables, future studies should consider gathering a sample from a more general population in order to enhance the generalizability of the study Product type and
85 involvement situation scenarios should also be appropriately modified for replication or extension of the current study. Fo urth, although both the pretest an d manipulation check showed statistically signific ant difference between high and low involvement situations (manipulation 2) the difference was not large as it was intended. Narrow difference was also detected for partic ipant perceived valence of positive vs. negative consumer reviews. The process of manipulation should be re examined for future study to maximize the difference between different conditions Lastly, a s with any other studies adopting SEM, the r elationsh ips among constructs suggested by the current study are only causal inferences based on theories It does not confirm the perfect causality as indicated in the model. In spite of these limitations, the current study c ontributes to understanding the underly ing mechanism of how consumers are influenced by online consumer reviews As of y et, t here have been f ew studies conducted to understand how consumers form their attitude and behavioral intention in responding to WOM ; therefore there are still a lot of qu estions yet to be answered, and a lot of areas yet to be examined.
86 APPENDIX Consumer Review (Functional Positive)
87 Consumer Review (Functional Negative)
88 Consumer Review (Hedonic Positive)
89 Consumer Review (Hedonic Negative)
90 LIST OF REFERENCES Aaker, D., & Myers, J. G. (1982). Advertising management (2nd ed.). Englewood Cliffs, NJ: Prentice Hall. Aaker, D., Stayman, D., & Hagerty, M. (1986). Warmth in advertising: Measurement, impact and sequence effect. Journal of Consumer Research 12, 365 381. Ajzen, I., & Fishbein, M. (2005). The influence of attitudes on behavior. In D. Albarracin, B. T. Johnson & M. P. Zanna (Eds.), The handbook of attitudes (pp.173 221). Mahwah, NJ: Lawrence Erlbaum Associates. Anderson, J. & Gerbing, D. (1988). S tructural Equation Modeling in p ractice: A r eview and r ecommended Two Step a pproach. Psychological Bulletin 103(3), 411 423. Antonides, G. (1996). Psychology in economics and business Dordrecht / Boston / London: Kluwer Academic. Arndt, J. ( 1967). Role of product related conversations in the diffusion of a new product. Journal of Marketing 4 (August), 291 295. Arndt, J., & May, F. (1981). The hypothesis of a dominance hierarchy of information sources. Journal of the Academy of Marketing Scie nce 9(4), 337 352. Assael, H., & Kamins, M. (1989). Effects of appeal type and involvement on product disconfirmation: A cognitive response approach through product trial. Journal of the Academy of Marketing Science 17(3): 197 208. Bagozzi, R. P., & Burn krant, R E. ( 1979). Attitude organization and the attitude behavior relationship. Journal of Personality and Social Psychology 27 (June), 913 929. Bagozzi, R. P., Gopinath, M., & Nyer, P. U. (1999). The role of emotions in marketing. Journal of the Acade my of Marketing Science 27(2), 184 206. Bagozzi, R. P., & Phillips, L. W. (1982). Representing and testing organizational theories: A holistic construal. Administrative Science Quarterly 27(3), 459 489. Bang S. (2006). Online word of mouth effect by posi tive versus negative proportion, direction of the consumer review Master thesis, Seoul, South Korea: Yonsei University. Batra, R., & Ahtola, O. T. (1990). Measuring the hedonic and utilitarian sources of consumer attitudes. Marketing Letters 2(2), 159 17 0. Batra, R., & Stephens, D. (1994). Attitude effects of ad evoked moods and emotions: The moderating role of motivation. Psychology and Marketing 11(3), 99 215. Bayus B. (1985). Word of mouth: The indirect effects of marketing efforts. Journal of Adverti sing Research 25 (3): 31 40.
91 Bickart B., & Schindler, R. M. (2001). Internet forums as influential sources of consumer information. Journal of Interactive Marketing 15(3), 31. Biswas D., Biswas, A., & Das, N. (2006). The differential effects of celebrity and expert endorsement on consumer risk perceptions. Journal of Advertising 35(2), 17 31. Bone P. F. (1992). Determinants of word of mouth communication during product consumption. Advances in Consumer Research 19(1), 579 583. Breazeale, M. (2009). Wor d of mouse. International Journal of Market Research 51(3), 297 318. Breckler, S. J. (1984). Empirical validation of affect, behavior, and cognition as distinct components of attitude. Journal of Personality and Social Psychology 47(6), 1191 1205. Briste r, J. M. (1991). Word of mouth communication and their effects in consumer network. Advances in Consumer Research 18, 155 169. Brown, J. J., & Reigen, P. (1987). Social tie and word of mouth referral behavior. Journal of Consumer Research 14, 350 362. Br own, S. P., & Stayman, D. M. (1992). Antecedents and consequences of attitude toward the ad: A meta analysis. Journal of Consumer Research 19(1), 34 51 Buck, R. (1988). Human motivation and emotion (2nd ed.). Oxford, England: John Wiley & Sons. Chaffee, S. H. (1979). Man media vs. Interpersonal channels: The synthetic competition. Paper presented at the Speech Communication Association Convention : San Antonio, TX. Chatterjee, P. (2001). Online review: Do consumers use them?. Advances in Consumer Research 29, 129 133. Chen, Y., & Xie, J. (2008). Online consumer review: Word of mouth as a new element of marketing communication mix. Management Science 54(3), 477 491. Cheung, C., Matthew, M., Lee, K., & Rabjohn, N. (2008). The impact of electronic word of mo uth: The adoption of online opinions in online customer communities. Internet Research 18(3), 229 247. Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: Online book reviews. Journal of Marketing Research 43(3), 345 354. Chiang K., & Dholakia, R. (2003). Factors driving consumer intention to shop online: An empirical investigation. Journal of Consumer Psychology 13, 177 183.
92 Chiu, H., Hsieh, Y., & Kao, C. (2005). Website quality and customer's behavioral intention: An explorat ory study of the role of information asymmetry. Total Quality Management & Business Excellence 16(2), 185 197. Clore, G. L., & Schnall, S. (2005). The influence of affect on attitude, In D. Albarracin, B. T. Johnson & M. P. Zanna (Eds.), The handbook of a ttitudes (pp. 437 489). Mahwah, NJ: Lawrence Erlbaum. Cohen, J. & Cohen, P. (1983). Applied multivariate regression/correlation analysis for the behavioral sciences. (2 nd ed.). London: Lawrence Erlbaum. C one Communications (2011). Online influence trend t racker Ret rieved February 1, 20 12 from http://www.coneinc.com/research/index.php Corey, S. M. (1937). Professional attitudes and actual behavior. Journal of Educational Psychology 29, 271 280. d'Astous, A., & Touil, N. (1999). Consumer evaluations of mo vies on the basis of critics' judgments. Psychology & Marketing 16(8), 677 694. Ding, L., Velicer, W. F., & Harlow, L. L. (1995). Effects of estimation methods, number of indicators per factor and improper solutions on structural equation modeling fit ind ices. Structural Equation Modeling 2(2), 119 143. Eagly, A. H., & Chaiken, S. (1993). The Psychology of Attitudes Fort Worth, TX: Harcourt Brace Jovanovich. Easwar, I. (1993). Anatomy of green advertising. Advances in Consumer Research 20, 494 501. Erve lles, S. (1998). The role of affect in marketing. Journal of Business Research 42, 199 215. Faber, R., & O'Guinn, T. (1984). Effect of media advertising and other sources on movie selection. Journalism Quarterly 61(2), 371 377. Fazio, R. H. (1986). How d o attitudes guide behavior?. In R. M. Sorrentino & E. T. Higgins (Eds.), The handbook of motivation and cognition: Foundations of social behavior (pp.204 243). New York, NY: Guilford Press. Fishbein, M. (1963). An investigation of the relationships between belfis about an object and the attitude toward the object. Human Relations 16, 233 230. Fishbein, M. (1965). Prediction of interpersonal preference and group member satisfaction from estimated attitudes. Journal of Personality and Social Psychology 1, 6 63 667. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research Reading, MA: Addison Wesley. Fiske, S. T. & Taylor, S. E. (1984). Social cognition Reading M A : Addison Wesley.
93 Forrester.com (200 5). Retrieved September 16, 2008, from Y. Lee. C Generation marketing trend. CJ Magazine at http://www.lgeri.com. Gardyn, R. (2002). How does this work?. American Demographics December 2002/January 2003, 18 19. Goldsmith, R., & Horowitz, D. (2006). Measu ring motivations for online opinion seeking. Journal of Interactive Advertising 6(2), 1 16. Gruen, T. W., Osmonbekov, T., & Czaplewski, A. (2006). eWOM: The impact of customer to customer online know how exchange on customer value and loyalty. Journal of Business Research 59, 449 456. Hansen. F., & Christensen, S. R. (2007). Emotions, advertising and consumer choice Gylling, Denmark: Copenhagen Business School Press. Havlena, W. J., & Holbrook, M. B. (1986). The varieties of consumption experience: Compa ring two typologies of emotion in consumer behavior. Journal of Consumer Research 13(3), 394 404. Haywood, K. (1989). Managing word of mouth communication. Journal of Service Marketing 3(2), 55 72. Hennig Thurau, T., Gwinner, K., Walsh, G., & Gremler, D. (2004). Electronic word of mouth via consumer opinion platforms: What motivates consumers to articulate themselves on the Internet?. Journal of Interactive Marketing 18(1), 38 50. Hennig Thurau, T., & Walsh, G. (2003). Electronic word of mouth: motives for and consequences of reading customer articulations on the Internet. International Journal of Electronic Commerce 8(2), 51 74. Holbrook, M. B. (1978). Beyond attitude structure: Toward the informational determinants of attitude. Journal of Marketing Re search 15, 545 556. Holbrook, M. B., & Batra, R. (1987). Assessing the role of emtions as mediators of consuemr refsponsese to advertising, The Journal of Consumer Research 14(3), 404 420. Holbrook, M. B., & Hirschman, E. C. (1982). The experiemental asp ects of consumption: Consumer fantasies, feelings, and fun. The Journal of Consumer Research 9, 132 140. Homer, P. M. (2006). Relationships among ad induced affect, beliefs, and attitudes. Journal of Advertising 35(1), 35 51. H o rst, P. (1965). Factor an alysis of data matrices New York: Holt Rhinehart and Winston. Hoyer, W., & MacInnis, D. (2007). Consumer behavior (4th Ed.). Boston, MA: Houghton Mifflin Company.
94 Hu, N., Liu, L., & Zhang, J. (2008). Do online reviews affect product sales? The role of rev iewer characteristics and temporal effects. Information Technology & Management 9(3), 201 214. Hung, K., & Li, S. (2007). The influence of eWOM on virtual consumer communities: Social capital, consumer learning, and behavioral outcomes. Journal of Adverti sing Research 47 (4), 485 495. Izard, C. E. (1991). The psychology of emotions New York, NY: Plenum Press. Jin, C. (2006). Animated commercials' effects on low effort routes to persuasion: Classical conditioning approach Doctoral dissertatio n, Universit y of Florida, Gainesville, FL. Joreskog, K. G. (1971). Statistical analysis of sets of congeneric tests. Psychometrika 36(4), 109 133. Jun, J., Cho, C., & Kwon, H. (2008). The role of affect and cognition in consumer evaluations of corporate visual identi ty: Perspectives from the United States and Korea, Brand Management 15(6). 382 398. Jung, N. (2006). The effects of WOM on the consumer product choice and redelivery in the online and offline environments Master thesis, Yonsei University, Seoul, South Ko rea. Kane, R. (1985). Free will and values Albany, NY: State University of New York Press. Katz, D., & Stotland, E. (1959). A preliminary statement to a theory of attitude structure and change. In S. Koch (Ed.), Psychology: A study of a science Formulati ons of the person and the social context (Vol. 3. pp. 423 475). New York, NY: McGraw Hill. Katz, E., & Lazarsfeld, P. (1955). Personal influence; The part played by people in the flow of mass communications Glencoe, IL: Free Press. Kempf, D. (1999). Attit ude formation from product trial: Distinct roles of cognition and affect for hedonic and functional products. Psychology and Marketing 16(1), 35 50. Kempf, D., & Robert, E. S. (1998). Consumer processing of product trial and the influence of prior adverti sing: A structural modeling approach. Journal of Marketing Research 35(3), 325 338. Kerlinger, F. N. & Pedhazer, E. J. (1973). Multiple regression in behavioral research New York: Holt, Rhinehart and Winston. Kim, J., Lee, J., & Ragas, M. (2011). Explor ing eWOM in o nline c onsumer r eviews: Experience v ersus s earch g oods W eb J ournal of M ass C ommunication R esearch 32 (May)
95 Kim, J., & Morris, J. D. (2007). The power of affecitve response and cognitive structure in product trial attitude formation. Journal of Advertising 36(1), 95 106. Krugman, H. (1966). The impact of advertising involvement. Public Opinion Quarterly 30(4), 583 596. Kukuk, C. R. & Baty, C. F. (1979). The misuse s of multiple regression with composite scales obtained from factor scores. Ed ucational and Psychological Measurement 39, 277 290. Lang, P. J. (1985). The cognitive psychophysiology of emotion: Anxiety and the anxiety disorders Hillsdale, NJ: Lawrence Erlbaum. Lang, P. J. (1995). The emotion probe Studies of motivation and atten tion. American Psychologist 50(5), 372 385. LaPiere, R. T. (1934). Attitudes vs. Actions. Social Forces 13, 230 237. Lazarus, R. S. (1982). Thoughts on the relations between emotion and cognition. American Psychologist 37, 1019 1024. Lazarus, R. S. (198 4). On the primacy of cognition. American Psychologist 39, 124 129. Leigh, J., Zinkhan, G. M., & Swaminathan, V. (2006). Dimensional relationships of recall and recognition measures with selected cognitive and affective aspects of print ads. Journal of Ad vertising 35(1), 105 122. MacKenzie, S. B., & Lutz, R. J. (1989). An empirical examination of the structural antecedents of attitude toward the ad in an advertising pretesting context. Journal of Marketing 53(2), 48 65. MacKenzie, S. B., Lutz, R. J., & B elch, G. E. (1986). The role of attitude toward the Ad as a mediator of advertising effectiveness: A test of competing explanations. Journal of Marketing Research 23(2), 130 143. Madhavaram, S. R., & Laverie, D. E. (2004). Exploring impulse purchasing on the Internet. Advances in Consumer Research 31, 59 66. Marks, L. J., & Michael, A. K. (1988). The use of product sampling and advertising: Effects of sequence of exposure and degree of advertising claim exaggeration on consumers' belief strength, belief c onfidence, and attitudes. Journal of Marketing Research 25(3), 266 281. Mehrabian, A., & James A. R. (1974). An approach to environmental psychology Cambridge, MA: MIT Press. Michael, Wes, & Morris, Jon (2004). Incorporating emotions into healthcare segm entation research. Presented at the Advertising Research Foundation (ARF ) Conference Fall 2004.
96 Mischel, W. (1996), From good intentions to willpower. In P. Gollwitzer and J. Bargh (Eds.), The psychology of action (pp. 197 218). New York: Guilford Press. Mittal, B (1989) Measuring Purchase Decision Involvement Psychology and Marketing 6 (Summer), 147 162. Morris, J. D. (1995). Observations: SAM: The Self Assessment Manikin An efficient cross cultural measurement of emotional response. Journal of Adverti sing Research November/December, 63 68. Morris, J. D., Klahr, N., Shen, F., Villegas, J., Wright, P., He, G., & Liu, Y. (2008). Mapping a multidimensional emotion in response to television commercials. Human Brain Mapping Published online Feb. 19, 2008 a t www.interscience.wiley.com. Morris, J. D., Woo, W., Geason, J., & Kim, J. (2002). The power of affect: Predicting intention. Journal of Advertising Research 42(3), 7 17. Muehling, D. D. (1987). An investigation of factors underlying attitude toward adve rtising in general. Journal of Advertising 16(1), 32 40. Nelson, P. (1974). Advertising as information. Journal of Political Economy 82(4), 729 755. Nunnally, J. C. (1978). Psychometric Theory (2nd Ed.). New York: McGraw Hill. Oh, S. (2005). Message anal ysis of the Internet bulletin board as word of mouth communication space of a movie according to relationship with movie types and determinants of box office performance Master thesis, Yonsei University, Seoul, South Korea. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning Urbana, IL: University of Illinois Press. Park, C., & Chung, S. (2006). A content analysis of online consumer reviews in Korea. Advertising Research 70, 91 118. Park, D., & Lee, J. (2008). eWOM overloa d and its effect on consumer behavioral intention depending on consumer involvement, Electronic Commerce Research and Application 7. 386 398. Park, D., Lee, J., & Han, I. (2007). The effect of on line consumer reviews on consumer purchase intention: The m oderating role of involvement. International Journal of Electronic Commerce 11(4), 125 148. Perloff, R. M. (2008). The dynamics of persuasion: Communication and attitudes in the 21st century New York, NY: Lawrence Erlbaum Associates.
97 Peterson, R., Balas ubramanian, S., & Bronnenberg, B. (1997). Exploring the implications of the Internet for consumer marketing. Journal of the Academy of Marketing Science 25(4), 329 346. Petty, R. E., & Cacioppo, J. T. (1981). Attitudes and persuasion: Classic and contempo rary approaches Boulder, CO: Westview Press. Petty, R. E., Unnava, H. R., & Strathman, A. J. (1991). Theories of Attitude Change. In T. S. Robertson & H. H. Kassarjian (Eds.), Handbook of consumer behavior (pp.241 280). Englewood Cliffs, NJ: Prentice Hall Pham, M., Cohen, J. B., Pracejus, J. W., & Hughes, G. D. (2001). Affect monitoring and the primacy of feelings in judgment. Journal of Consumer Research 28(September), 167 188. Reinstein, D., & Snyder, C. (2005). The influence of expert reviews on consu mer demand for experience goods: A case study of movie critics. Journal of Industrial Economics 53(1), 27 51. Reisenzein, R. (1986). A structural equation analysis of Weiner's attribution affect model of helping behavior. Journal of Personality and Soci al Psychology 50, 1123 1133. Richins, M. (1983). Negative word of mouth by dissatisfied consumers: A pilot study. Journal of Marketing 47(1), 68 78. Richins, M. (1984). Word of mouth communication a negative information. Advances in Consumer Research 11 (1), 697 702. Richins, M. (1997). Measuring emotions in the consumption experience. Journal of Consumer Research 24(2), 127 146. Ritchins, M., & Root Shaffer, T. (1988). The role of enduring involvement and opinion leadership in consumer word of mouth: an implicit model made explicit. Advances in Consumer Research 15(1), 35 43. Rosen, E. (2000). The anatomy of buzz (1st ed.). New York, NY: Doubleday. Rosenberg, M. J. (1968). Hedonism, inauthenticity, and other goads toward expansion of a consistency theor y. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg & P. H. Tannenbaum (Eds.), Theories of cognitive consistency: A source book Chicago: IL, Rand McNally. Rummel, R. J. (1970). Applied factor analysis Evanston, IL: Northwestern University Press. Russell, J. A., & Mehrabian, A. (1977). Evidence for a three factor theory of emotions. Journal of Research in Personality 11, 273 294.
98 Samson, A. (2006). Understanding the buzz that matters: Negative vs positive word of mouth. Internat ional Journal of Market Research 48(6), 647 657. Schindler, R., & Bickart, B. (2005). Published word of mouth: Referable, consumer generated information on the Internet. In C. P. Haugtvedt (Ed.), Online consumer psychology: Understanding and influencing c onsumer behavior in the virtual world Mahwah, NJ: Lawrence Erlbaum Associates Publishers. Sen, S., & Lerman, D. (2007). Why are you telling me this? An examination into negative consumer reviews on the web. Journal of Interactive Marketing 21(4), 76 94. Shimp, T. A. (1981). Attitude toward the ad as a mediator of consumer brand choice. Journal of Advertising 10(2), 9 15. Shiv, B., & Fedorikhin, A. (1999). Heart and mind in conflict: Interplay of affect and cognition in consumer decision making. Journal o f Consumer Research 26(December), 278 282. Shiv, B., Britton, J. A., & Payne, J. (2004). Does elaboration increase or decrease the effectiveness of negatively versus positively framed messages. Journal of Consumer Research 31, 199 208. Smith, R. E. (199 3). Integrating information from advertising and trial: Processes and effects on consumer response to product information. Journal of Marketing Research 30(2), 204 219. Tibbetts, P. (2001). Ho w to spread the word of mouse Silt, CO: Lunatic Fringe Publish ing. Triandis, H. C. (1971). Attitude and attitude change New York, NY: Wiley. Tsal, Y. (1985). On the relationship between cognitive and affective processes: A critque of Zajonc and Markus. Journal of Consumer Research 12(December), 358 362. Verma, R., Davatzikos, C., Loughead, J. Indersmitten, T., Hu, R., Kohler, C., Gur, R. E., & Gur, R. C. (2005). Quantification of facial expressions using high dimensional shape tranformations. Journal of Neuroscience Methods 141(1), 61 73. Wang, A. (2005). The effec ts of expert and consumer endorsements on audience response. Journal of Advertising Research 45(4), 402 412. Whyte, W. H. Jr. (1954). The web of word of mouth. Fortune 50 (November), 140. Wimmer, R., & Dominick, J. (1997). Mass media research: An introdu ction (5th ed.). Belmont, CA: Wadsworth Publishing Company. Woods, W. (1960). Psychological dimensions of consumer decision. Journal of Marketing 24(3), 15 19.
9 9 Yan g Z., & Fang, X. (2004). Online service quality dimensions and their relationships with sat isfaction: a content analysis of customer reviews of securities brokerage services. International Journal of Service Industry Management 15(3), 302 319. Yoon, Y. (2000). On line political discussion and deliberative democracy. Journal of Korean Broadcasti ng and Telecommunication 14(2), 109 119. Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology 9(2), 1 28. Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologis t 39, 117 123. Zajonc, R. B., & Markus, H. (1982). Affective and cognitive factors in preferences. Journal of Consumer Research 9(2), 123 131. Zeitlin, D. M., & Westwood, R. A. (1986). Measuring emotional responses. Journal of Advertising Research 26(5) 34 44.
100 BIOGRAPHICAL SKETCH J insoo Kim was born in South Korea in 1971. He received his Bachelor of Arts degree in Russian Language and Literature from Korea University (Seoul, South Korea) in 1996. He earned his master s degree from University of Mis souri Columbia in 2000. After six years of advertising industry experience at a multi national company (McCann Erickson) in Korea, he came back to the United States to pursue his Ph.D. degree in a dvertising. During his stay at the University of Florida, he conducted several research projects on eWOM and emotion under the guidance of Dr. Jon D. Morris. In fall, 2011, he accepted a position and has been working as an a ssistant p rofessor of c ommunication ( a dvertising) at the Rhode Island College, Providence, RI. He received his Ph.D. from the University of Florida in the spring of 2012.