Can spending money make some of us happier? : the role of maximization in post-purchase satisfaction for three types of goods and the impact of online reviews

(1)

Can spending money make some of us happier?

The role of maximization in post-purchase satisfaction for three

types of goods and the impact of online reviews.

Master’s Thesis

Author : H.R.U. Rinne BSc (11578378) Supervisor : T. Dudenhöffer

Date of submission: 26/01/2018

UvA Amsterdam Business School

(2)

Statement of originality

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

(3)

Abstract

The aim of this research is to study the relationship between maximizing and satisfaction, further to examine the impact type of good and online reviews have on this relationship. The basic assumptions behind the research question is built upon the theories of maximizing, cognitive dissonance, and the literature on trichotomy of goods and online customer reviews. A preliminary study was conducted to uncover how people characterize goods. The main study is conducted with an online survey, and it measures the tendency to maximize plus the effect this has on the satisfaction post-purchase when looking at a physical product, an experience and a credence good. Additionally, the moderating effect of online customer reviews is examined. The results show that there is no significant effect of maximizing on satisfaction post-purchase, which is in contrast with expectations. Further, no significant moderating effects of online reviews or type of good were found on the relationship between maximizing and satisfaction. A significant main effect was found of product type on satisfaction and online reviews on satisfaction. Thus, the study demonstrates the need for marketers to consider the characteristics of the sold good in order to achieve the desired result of satisfied consumers. Additionally it highlights the importance and impact of online consumer reviews in the modern marketplace.

Key words_{: Maximizing, Decision making, consumer behaviour, Satisfaction, Cognitive} Dissonance, Commitment, Choice, Online consumer reviews, Physical product, Experience, Credence good

(4)

Acknowledgements

This master’s thesis brings me to the end of my student life, and it is the result of hard work. Nevertheless, the process would not have been possible without the help of supervision, friends and family.

I would like to thank my supervisor Tina Dudenhöffer, for I could always count on advice from her when I was in doubt. She critically reviewed my work and engaged in discussing my topic, always giving me a fresh perspective and new ideas.

In addition, this thesis could not have been completed without the encouragement from my parents back in Finland, who supported my decision of doing a master at UvA and over the phone gave me the push I needed throughout the year. Lastly, I want to thank my roommate Sophie Loos for the many late nights we spent together working on our researches, motivating and helping each other.

(5)

List of tables

Table 1. Results pre-study……….……….……….……….………. 30 Table 2. Reliability Analysis DV, IV and moderators ……….….………39 Table 3. Correlation Matrix……….……….……….……….………….. 41 Table 4. Summary of PROCESS regression model results (model 2, Jeans)……….….………. 45 Table 5. Summary of PROCESS regression model results (model 2, Holiday)……….……….. 46 Table 6. Summary of PROCESS regression model results (model 2, Reparation)……… 47 Table 7. Summary of results……….……….……….……….…………..48

List of figures

Figure 1. The Two-Component Model of Maximization ……….…….. 14 Figure 2. Conceptual model ……….……….……….……….…………. 26 Figure 3. PROCESS model 2, Double Moderation ……….……… 44

List of Abbreviations

CD……….……….……….………. Cognitive Dissonance DV……….……….……….………. Dependant Variable IV……….……….……….……….... Independent Variable MAX……….……….……….……….. Maximizing OnRev ……….……….……….………. Online Reviews SAT……….……….……….………... Satisfaction

(8)

1. Introduction

Think about the previous time you were purchasing a pair of jeans; did you have a hard time selecting the perfect pair? Maybe you even made the purchase, but regretted it and decided to return them the next day because you were not fully satisfied. Now think about your last holiday abroad. If we are to believe past research (e.g. Carter & Gilovich, 2014), you did not regret this purchase, but were quite satisfied with it. Further, your decision was most likely affected by the large amount of available online reviews by other customers. In fact, research shows that up to 86% of consumers under the age of 45 consider online reviews as an essential part of their purchase process (Anderson & Magruder, 2012). Does this mean that you automatically would be happier spending your disposable income on experiences rather than products? Not necessarily, as there are both personal and contextual factors influencing this.

Past research in the area of decision making has shown that the task of making choices is more exhaustive to certain types of people, due to personal variations (e.g. Gourville and Soman, 2005; Hayes, 2013). Especially the trait of _{maximizing, the goal of choosing the best, has} attracted a large amount of research in the recent few years (e.g. Simon, 1965; Schwartz _{et al.,} 2002; Cheek & Schwartz, 2016). The researchers describe how a person with high maximization tendencies experiences more difficulties when making choices, and as a result further suffers more from the negative side of free choice, such as regret and dissatisfaction post-choice. The trait of maximizing can make everyday life more difficult, as people in the modern western countries are faced with an increasing amount of choices to be made every day. Choice is a highly valued right for us, and it empowers the consumer to act out in line with personal

(9)

preferences. However, there seems to be a negative side to the free choice as well. Although consumers have reported to feel more free with more options (e.g., Reibstein, Youngblood & Fromkin, 1975), the extensive amount of research shows that free choice and more options might not always lead to better outcomes and more satisfaction (e.g., Chan, 2015; Haynes, 2009; Iyengar & Lepper, 2000; Scheibehenne Greifeneder _{& Todd, 2010). The phenomena of decrease} in motivation, or even inability, to make a choice due to a large amount of options has been labelled ”_{choice overload effect”} (Diehl & Poynor 2007; Iyengar and Lepper 2000, Mogilner, Rudnick, and Iyengar2008) or ”_{paradox of choice” (} Schwartz, 2004).This effect has shown to have consequences such as experiences of regret, dissatisfaction with choice, unhappiness and mental discomfort, especially for a person with high tendency to maximize (Dar-Nimrod _{et al.,} 2009; Gourville and Soman 2005, Haynes, 2009 ; Iyengar & Lepper, _{2000; Schwartz et al.} 2002). This poses a challenge for companies that increasingly are competing for satisfied and loyal customers.

Moreover, previous research has noticed that physical products are related to to more choice making difficulties and dissatisfaction compared to experiences. This is because physical products can be compared in a larger extent, prior to purchase (Nelson, 1970; Carter & Gilovich, 2014). Experiences on the other hand are difficult to evaluate prior to purchase, further become a part of oneself after consumption, which in general results in higher satisfaction. What has not yet been researched is how decisions regarding credence goods, that are difficult to evaluate, affect satisfaction (eg. Hackley, 2006; Sparks_{et al. 2012,; Carter & Gilovich, 2014).}

(10)

As online reviews are becoming an increasingly important part of our purchase journey, research has noted that they are influencing the purchase process. A study by Anderson & Magruder (2012) shows that 86% of respondents think reviews are essential to them when making a purchase decision. Research further shows a positive relationship between reading an online review and increased likelihood of purchase (eg. Chen, Wu & Yoon 2004, Askalidis & Malthouse 2016). Naturally there are different types of online reviews, both written by the company itself (comparable with advertising), reviews written by experts and reviews written by fellow consumers. As research shows that consumers trust fellow consumer’s reviews the most ( Nielsen, 2012) these are the types of reviews the current paper is discussing. The current research is interested in looking at how the existence of online consumer reviews possibly facilitate the decision making process and satisfaction with the choice made.

There is not much empirical research looking into the effect the different types of goods have on satisfaction with the choice, and how the decision making process differs depending on the tendency to maximize. Further, there is a need of looking into how the increasingly important online reviews facilitate the decision making among consumers, especially those with a tendency to maximize. Building onto these topics, this paper aims to contribute to that gap by answering the research question ‘In what way does the tendency to maximize influence the post-purchase satisfaction? How do type of good and online consumer reviews moderate that relationship?’

To answer the research question, primary data was collected for this study through two online surveys among consumers living in western countries. First, a preliminary study was conducted to determine how people categorize goods. After this a main study was realized. First it measured the tendency to maximize. After, the respondents were asked to think of the previous

(11)

time they purchased the following goods; jeans, holiday and reparation of phone. Post purchase satisfaction was assessed by measuring the level of cognitive dissonance for each good, together with questions regarding the consultation of online reviews prior to the particular purchase. It was expected that a person with a high tendency to maximize in general would be less satisfied with the purchases, in line with previous theory (e.g. Lai, 2011; Sparks _{et al., 2012). Further, the} type of good was expected to influence the relationship, so that satisfaction would be higher for reparation of phone and holiday compared to jeans, as a product is more easy to compare than experiences and credence goods, increasing the tendency to maximize.

This paper contributes to the field of consumer behaviour in various ways. Firstly, it responds to the request by Cheek & Schwartz (2016) to explore possible moderators of maximization, that in this research are type of good and impact of online reviews. Moreover, maximization has not been studied in context with credence goods that are difficult for consumers to evaluate, even post consumption. The paper looks into the differences between satisfaction with physical products, experiences and credence goods. In addition, as online reviews play a crucial role in modern purchase decisions, the impact they have on satisfaction are looked into. Thus, the main contribution of this paper is including the moderating effects of product type and online reviews on the relationship between maximizing and satisfaction post-purchase. A better understanding of how these two moderators influence satisfaction with purchase gives companies a chance to reflect over how they choose to market different types of goods for different types of consumers. In addition, knowledge about the effect online reviews have on the purchase intent and satisfaction should be taken into consideration in order to achieve full potential of directing a doubting consumer to making the purchase.

(12)

The paper is structured as follows; firstly a theoretical framework presents the relevant literature related to the areas of interest. The chapter ends with the presentation of the developed conceptual model and hypotheses. After, the empirical part of the research is described, presenting the preliminary study, methodology, data collection and analysis. The paper ends with a discussion of the results and the implications they have, both for theory and management in marketing. Lastly, limitations and suggestions for future research are given.

2. Theoretical Framework

The following chapter will provide an extensive overview of the literature on maximization and its relationship to cognitive dissonance and satisfaction. The literature review further explores how online reviews and three different types of goods have an effect on this relationship. First the theory of maximization will be discussed. After, the outcomes of maximization are depicted, the theory of cognitive dissonance is presented, which in this research is used as a measurement of post purchase satisfaction. Third, the three types of consumer goods and the concept of online reviews are presented, after which they will be linked to the relationship of maximization and post purchase satisfaction. The chapter ends with a conceptual model illustrating the current research and hypothesized outcomes.

2.1 Theory of Maximization

Herbert Simon (1955, 1956) was one of the first to bring up the ideas of maximizing and satisficing. The concepts were born as a result of his criticism on the _{Rational Choice Theory by} von Neumann & Morgenstern (1944) .The Rational Choice Theory firstly states that we all have

(13)

well-ordered preferences we tend to follow. In a decision making scenario we compare all available options and order them on a scale of preference, value and/or utility. After the comparison we will choose the option that is the most preferable for us, we therefore maximize our preferences. (von Neumann & Morgenstern, 1944)

About a decade later, taking the human cognitive limitations into account, Simon (1956) argued that people in fact are not able to make the “best possible decision” as explained by the Rational Choice Theory. This was argued to be due to the limited information process ability of humans and the possibly endless amount of options to choose from. Instead of maximizing, people satisfice. This means settling for the first option we come across that satisfy our own decision goals and preferences. This is not necessarily the objectively best option. Simon describes his insight as following:

. . . from the examination of the postulates of these economic models it appears probable that, however adaptive the behavior of organisms in learning and choice situations, this adaptiveness falls short of the ideal of “maximizing” postulated in economic theory. Evidently, organisms adapt well enough to “satisfice”; they do not, in general, “optimize” _{(Simon, 1956,} p. 129).

With this insight Simon set the status quo for people to satisfice, not maximize as the Rational choice theory proposed. According to Simon’s view a person that satisfices, will evaluate goods until his or her own threshold of standard is exceeded. The first good to pass this threshold is chosen. He noted that a satisficers standards are nevertheless not automatically lower or more fixed than those of a maximizer. If a good of higher value is encountered later on by chance, the satisficer can change the choice. The choices are thus dynamic and gradually

(14)

increasing, so that they eventually come close to those of maximization. However, the specific goal of maximization itself is unachievable and not pursued by people. This is the underlying motive that differentiates a satisficer from a maximizer; a person who satisfices does not seek for the best available option, but an option that is good enough for him/herself. (Simon, 1956)

Upon this original research by Simon (1955, 1956), Schwart _{z et al. (2002) proposed and} further developed the theory of maximization as we know it today _. _{Schwartz et al. (2002)} suggested that people have different goals when making decisions so that some people satisfice; making choices that meet their standards, and others maximize; striving to make the best possible choice. Schwartz _{et al. (2002) further developed the Maximization Scale (MS) to} measure the degree to which individuals maximize. Most research on maximization has since then used this scale as a measure. Nevertheless, the MS has later been criticized for insufficient validity and reliability scores, and as a result up to 11 other measures and definitions of maximization has been published since the original MS in 2002 (see Nenkov _{et al., 2008; Diab et} al_{., 2008; Lai, 2010; Turner et al., 2012; Weinhardt et al., 2012; Mikkelson & Pauley, 2013;} Richardson _{et al., 2014; Ma & Roese, 2014; Dalal et al., 2014; Misuraca et al., 2015; Cheek &} Schwartz, 2016).

The many different publications have naturally lead to some confusion and contradiction when it comes to both the definition and measurement of maximization. Cheek & Schwartz (2016) tried to address this particular issue in a recently published article with a goal of clarifying the concept of maximization and evaluating the different scales that have been used for its measurement. Cheek & Schwartz (2016) summarizes six separate constructs that have been proposed to make up or relate to maximization in the past research; _{high standards, alternative}

(15)

research, decision difficulty, satisficing, regret and minimizing_{. According to the authors a} substantial weakness in the existing literature is the lack of a coherent and complete definition of the concept itself. Further they criticize how some of the later researchers’ views even have been in contrast to Simon’s (1955) original theory. This has lead to many imperfect ways to measure maximization. In their definition Cheek & Schwartz (2016) wanted to highlight the distinguishment between the actual components of maximization and the other constructs (such as regret and decision difficulty) that best are seen as outcomes or causes, but not components. (Cheek & Schwartz, 2016)

The initial conceptualization of maximization by Simon (1955) comprised two things, a goal of finding the best option and a strategy to get there. In their recent definition of maximization Cheek & Schwartz (2016) thus proposes a two-part construct to most accurately illustrate the way in which maximization should be conceptualized and measured. The Two-component model of maximization can be seen in Figure 1.

(16)

The first component in the model is the _{goal of a maximizer to choose the best, as seen on the} left side in Figure 1. This component reflects the ‘having high standards’ aspect of maximization, as highlighted by Simon (1955) among others. According to Cheek & Schwartz (2016) this goal of choosing the best might even be the definitional component in itself. The second component concerns the maximization strategy of alternative search, as can be seen in the right side of Figure 1. This is the maximization strategy in which the alternative options are sought and compared to each other. The model thus combines both seeking (information about) alternatives and comparing them. The authors explain how the alternative seeking is more relevant for decision making scenarios with a large amount of option, whereas the act of seeking out more information about the options available and using that information in the comparison is more relevant when there are few options. (Cheek & Schwartz, 2016)

In conclusion, to be a maximizer an individual must “... _{pursue the goal of choosing the} best through the strategy of alternative search, as opposed to through the use of other strategies._{” (Cheek & Schwartz 2016, p.136).} It is finally important to distinguish between having high standards and wanting the absolute best, as having high standards alone is not a criteria enough for maximization. A satisficer and maximizer can in fact have equally high standards, but it is only the maximizer who keeps comparing and searching for options after the personal standard is met. (Cheek & Schwartz, 2016)

As a response to the dilemma of how to best measure maximization Cheek & Schwartz (2016) suggests the combination of two scales in line with their model (Figure 1); the Seven Maximization Tendency Scale (MTS-7) to measure the ‘goal of choosing the best’ ( Dalal _{et al.,} 2015) and the alternative search subscale of the MS-S by Nenkov _{et al. (2008) to measure the}

(17)

‘maximization strategy’. These two scales are used in the research of this paper to determine the tendency to maximize amongst respondents.

2.1.1 Outcomes of Maximization

A maximizer has by definition a goal of only going for the best alternative, which naturally can be seen in the objective outcome of the decisions. For example Iyengar, Wells, & Schwartz (2006) showed in a study how maximizers negotiated a 20% better starting salary than satisficers, still they reported to be less satisfied with the outcome. It thus seems like maximizers tend to feel worse about objectively speaking better results than _{satisficers. (Schwartz et al.,} 2002; Iyengar, Wells, & Schwartz, 2006)

Why maximizers feel less satisfied with their choice might be explained by the lower commitment to choice they have (Sparks _{et al., 2012).} Because they carefully consider every available option, they might never be certain whether or not they made the right decision, which undermines their commitment to a selected choice (Hackley, 2006; Sparks _{et al., 2012).} Interestingly it has also been shown that people with a high maximization tendency prefer to put themselves in situations with more options in order to avoid committing to choices and to keep their options open (Hackley, 2006). Finally, Lai (2011) found that a consequence of this lower commitment to a made choice is lower consumer loyalty, when comparing maximizers to satisficers.

(18)

2.2 Cognitive Dissonance Theory

Cognitive dissonance (CD) theory was first introduced by Festinger in 1957. CD describes how people strive for psychological consistency in order to be able to function properly mentally. In short, if a person holds contradictory thoughts at the same time the result is psychological inconsistency (Festinger, 1957). The theory explains how a pair of cognitions (feelings, attitudes, perceptions, motivations or behaviors) either are consonant (in line) or dissonant (opposing) with each other. If the cognitions are dissonant this causes mental discomfort, namely cognitive dissonance, that motivates a person to eliminate it. Although the theory of cognitive dissonance has been subject to quite some critique since its introduction, no researcher has up to this day been able to displace the original theory (Festinger, 1957; Harmon-Jones & Mills, 1999; Sparks et_{al., 2011;}Bolia, Jha and Jha, 2016).

Marketeers have been interested in using CD particularly to investigate the experienced dissonance in a purchase situation, as Bolia, Jha and Jha (2016) show in their meta study. When experiencing mental discomfort (Festinger, 1957), a consumer starts to look for information that is supportive and favorable for either making or not making the decision ( Bolia, Jha and Jha 2016). In a purchase situation cognitive dissonance mainly affects the behaviour after the purchase, although some research also has studied pre-purchase dissonance (Bolia, Jha and Jha, 2016). In a post-purchase situation cognitive dissonance can occur when a consumer finds information that contradicts their mental cognition, e.g. when she compares the purchased product to the other available alternatives (Powers and Jack, 2013). If the comparison is not in favour for the made decision, psychological discomfort might emerge as a result (Festinger, 1957; Elliot and Devine, 1994). Feeling associated to this might be anxiety, uncertainty or doubt

(19)

(Menasco and Hawkins, 1978) as well as regret (Insko and Schopler, 1972). If a consumer experiences post-purchase dissonance, they are likely to seek to reduce dissonance by e.g. returning the purchase (Gilovich & Medvec, 1995; Powers and Jack, 2013). However this is naturally not possible for all types of goods, such as experiences or credences, which will be discussed later.

It is worth to mention that there is some research pointing out how CD is not present in

every purchase situation, that there are certain conditions that must be met in order for cognitive dissonance to occur. First the decision must be important to the consumer. This can be because a lot of money or psychological cost is invested in the purchase decision. Second, the purchase decision must be voluntary. Finally the decision should be somewhat permanent. (Sweeny _{et al.,} 2000, Wilkins, Beckenuyte and Butt, 2016)

There are many researchers that have sought to developed techniques to measure cognitive dissonance, including both cognitive, psychological as well as behavioural measures (Wilkins, Beckenuyte and Butt, 2016; _{Bolia, Jha and Jha, 2016). A conclusion of the previous} research is that there is no clear and consistent solution for the measurement of the abstract construct, but in a purchase setting the most widely used scale is developed by Sweeney _{et al.} (2000). Their multidimensional scale measures both the emotional and cognitive aspects of dissonance post-purchase. The 22-item scale is used in this research to determine post-purchase dissonance amongst participants, as a measurement of satisfaction.

(20)

2.2.1 Cognitive Dissonance related to Maximization and post-purchase satisfaction

Cognitive dissonance post-purchase depicts the state of mind a consumer experiences after a purchase has been made. If the consumer experiences dissonance, a necessary criteria to reduce it is commitment to the made choice (Allen, 1964; Sweeney _{et al., 2000; Wilkins, Beckenuyte and} Butt, 2016). Thus, without commitment to the choice, chances are bigger that the person does not experience an urge to reduce dissonance, leaving them with high cognitive dissonance. People with higher maximizing tendencies are usually less committed to a choice they make in the first place as they are expected to keep searching for better alternatives (Lai, 2011). As a consequence, they are not enjoying the psychological benefits of the commitment, such as cognitive dissonance reduction through increasing their liking for the chosen option and/ or decreasing their liking for rejected options (Festinger, 1957; Hackley, 2006; Sparks _{et al., 2012).} This is expected to leave them less satisfied, as a result of the unfavourable feeling of cognitive dissonance. A person with a lower tendency to maximize is more likely to commit to his/her choice, and as a consequence of this reduce the cognitive dissonance that might occur. This person is ultimately more likely to be satisfied with the choice as it in his/her perception has been justified for. (Sparks _{et al., 2011; Lai, 2011)}

The level of experienced dissonance is thus expected to differ depending on the tendency to maximize, so that a person more likely to maximize is expected to experience more dissonance in general after a purchase decision, mainly due to the low commitment (e.g. Sparks et al., 2012). Post purchase dissonance is further argued to have a relation to satisfaction with the choice made, so that the more dissonance a person is experiencing, the less satisfied with the choice she is due to the feelings of uncertainty, doubt, regret and/or anxiety (e.g. Menasco and

(21)

Hawkins, 1978; Montgomery and Barnes, 1993; Insko and Schopler, 1972). As a maximizer is less likely to reduce dissonance the following hypothesis will be checked for in the research: H1: _{There is a negative relationship between tendency to maximize and satisfaction post} purchase.

2.3 Trichotomy of Consumer Goods

Nelson (1970) was the first to introduce the distinction between classification of goods. He explained that a good can be classified into two categories based on its characteristics, either search goods or _{experience goods. The significant difference between these categories is that the} quality of a search good can be determined prior to purchase, whereas the qualities of an experience good cannot (Nelson, 1970).

Darby & Karni (1973) later added a third category to the grouping of Nelson (1970), goods that were characterized by _{credence. The utility of a credence good is difficult to evaluate,} even after purchase and consumption. The credence good market is further characterized by an information asymmetry between seller and consumer; the consumer is never sure whether or not the experience/service they received actually was good. A prime example of a credence good is a medical treatment. There are however currently few large-scale experiments with credence goods, and the markets in which research has been conducted in are mostly limited to medical care and car reparation (Ford, Smith & Swasy, 1988; Dulleck, Kerschbamer and Sutter, 2011).

These three characteristics together form the so called trichotomy of goods; a continuum for evaluating goods that go from search goods (easy to evaluate), _{to experience goods} (moderately easy to evaluate) to _{credence goods (next to impossible to evaluate). On the}

(22)

continuum of characteristics, services, especially performed by professionals and specialists, usually fall into the credence good end as they are characterized by intangibility, non-standardization and inseparability, making them difficult to evaluate. Physical products usually fall closer to the search _{product end of the continuum as they are easier to evaluate.} (Zeithaml, 1981)

Type of good has previously shown to have an impact on the decision making (eg. Hackley, 2006; Sparks_{et al., 2012; Carter & Gilovich, 2014) The following section will discuss} the phenomena in more detail.

2.3.1 Type of good in relation to maximization in decision making

A study by Carter & Gilovich (2014) showed that material purchase decisions encourages a person to use a more of a maximization strategy whereas experiential purchases encourages that in a much less extent. Furthermore, there are studies showing that people in general tend to be more satisfied with experience purchases than with product purchases. This is argued to be due to the fact that experiences are more difficult to compare than material products. Thus, choosing between experiences is often easier than choosing between material products, as the later can be compared in a greater extent _{(Carter & Gilovich, 2010; Sparks et al., 2012). Interestingly,} material purchases are also more likely to cause post-choice regret, as they remain in their true form throughout time. Experiences on the other hand change form and become a part of the self, thus they cannot be undone or mentally exchanged (Carter & Gilovich, 2010).

(23)

As previously noted, material products encourages people to maximize to a greater extent, and intangible products, such as experiences and credences, do so in far less extent. This leads to the following assumptions:

H2: _{The effect of tendency to maximize on satisfaction is moderated by the type of good the} purchase decision is concerned with.

The line between experiential and physical goods can sometimes be blurry; for example a new television screen can either purely be a product, but when you watch a movie with your friend it can become an experience. Carter & Gilovich (2010) argues that the most important distinction is how the person herself categorizes _{a product. The more a person categorizes the} good as a physical product, the more difficult the decision making will be causing cognitive dissonance post-purchase. This leads for the following hypothesis:

H2a_{: The relationship between maximisation tendencies and satisfaction is moderated by} product type, so that this relationship is more negative for a physical product.

For a good the consumer characterizes as an experience however the decision making is expected to be easier, thus the cognitive dissonance post purchase is expected to be lower. The hypothesis reads as following:

H2b_{: The relationship between maximisation tendencies and satisfaction is moderated by} product type, so that this relationship is less negative for an experience.

What has not been researched up to this day is the effect a credence good has on this relationship. As a credence good is characterized by information asymmetry between seller and buyer in addition with evaluation difficulty even post consumption, it can be argued that a

(24)

purchase decision of a credence good would lead to low cognitive dissonance, as the purchase is permanent and non-comparable. Therefore following is hypothesized:

H2c_{: The relationship between maximisation tendencies and sat is moderated by product type, so} that this relationship is less negative for a credence good.

2.4 Online Consumer Reviews

Online consumer reviews (also called Electronic Word of Mouth) have increasingly become important in (e-)commerce and are a natural part of a consumer’s purchase journey. By sharing reviews online the future search costs of consumers are lowered, and previous literature has shown that too high search costs can prevent customers from making a purchase (Stiglitz, 1989). Consumer based online feedback mechanisms also generates trust, which in turn can alleviate information asymmetry between buyer and seller (Ba and Pavlou, 2002). User reviews are especially important in trust building between the customers (Askalidis and Malthouse, 2016). In a recent survey Anderson and Magruder (2012) report that 86% of the respondents (under the age of 45) say reviews are essential when making purchase decisions, and up to 30% of the shoppers report that they consult consumer reviews for every purchase they make. They further found that reviews is the second most impactful factor after the price on the purchase decision. The literature also shows evidence for a positive relationship between customer reviews and increased purchase likelihood online, and this effect applies for all types of reviews regardless of the valence or other characteristics (eg. Chen, Wu & Yoon, 2004; Askalidis & Malthouse, 2016). Previous studies have however shown that most online reviews are relatively positive (Godes and Mayzlin, 2003; Resnick and Zeckhouser, 2002; Schinder and Bickart, 2012). According to

(25)

naturally for those consumers who displays them. This is something the majority of online retailers are aware of, in fact all of the top- 10 U.S. online retailers provide consumers with user reviews for their products. It is thus seems beneficial for online retailers to display reviews, even though they would not have control over the content or valence of the reviews. (Chen, Wu & Yoon, 2004; Schinder and Bickart, 2012; Askalidis & Malthouse, 2016)

2.4.1 Online reviews related to cognitive dissonance and satisfaction

Online reviews have a positive impact on decision making and purchase intention in the sense that it can direct the decision maker in one or another way. If a consumer is faced with mental discomfort during or after a purchase situation he or she will look for information that is supportive and favorable for the decision that is or has been made (Festinger, 1957; Bolia, Jha and Jha, 2016). Therefore it is expected that the mere existence of reviews already will make the decision process easier, resulting in less cognitive dissonance. Further, the effect is expected to be larger for a person with more decision difficulties, in other words with high maximizing tendencies. This expectation is based on previous research by Sparks _{et al., (2011) reporting that} people with high maximizing tendencies are more affected by external feedback. If the feedback is in line with the made decision this increases their satisfaction because they feel they made the right choice, committing more to their decision and affording themselves benefits of dissonance reduction (Sparks _{et al., 2010). H3 will therefore be tested as following:}

H3_{: The effect of tendency to maximize on satisfaction is moderated by online reviews, such that} the relationship is less strong when a review is present in the decision making process.

(26)

As described earlier, reviews are in general more positive than negative. However, too much positivity might have an unfavorable effect on the credibility of the review. Moreover, Schindler and Bickart (2012) did not find a relationship between review value and the amount (or even existence) of negative evaluative statements. In fact, negative or mixed reviews are especially valued when ruling out alternatives in a decision making process (Schindler and Bickart, 2005), and further negative reviews are in some cases perceived as more informative, useful and accurate than positive ones (Sen and Lerman, 2007). In conclusion there does not seem to be a completely coherent opinion about the resonance of the review, however based on the research, a negative review is also likely to have a positive impact on the decision making (eg. Chen, Wu & Yoon, 2004; Schindler and Bickart, 2012; Askalidis & Malthouse, 2016) This is why the characteristics of the review(s) ultimately will not be taken into account in the current study, when measuring the moderating effect of online consumer reviews.

2.5 Conceptual model

Concluding from the previous literature review, there is a need for further research in the area of maximizing in a purchase situation and the effect this has on post-purchase satisfaction. It is thus motivated to explore how the trait of maximization relates to the experienced cognitive dissonance and to look at the moderating effects of product types and online reviews. A conceptual model as can be seen below in Figure 2 is developed for the study.

(27)

Figure 2._{Conceptual model}

H1_{: There is a negative relationship between tendency to maximize and satisfaction post} purchase.

H2: _{The effect of tendency to maximize on satisfaction is moderated by the type of good the} purchase decision is concerned with.

H2a_{: The relationship between maximisation tendencies and satisfaction is moderated by} product type, so that this relationship is more negative for a physical product.

H2b_{: The relationship between maximisation tendencies and satisfaction is moderated by} product type, so that this relationship is less negative for an experience.

H2c_{: The relationship between maximisation tendencies and sat is moderated by product type, so} that this relationship is less negative for a credence good.

(28)

H3_{: The effect of tendency to maximize on satisfaction is moderated by online reviews, such that} the relationship is less strong when a review is present in the decision making process.

The following chapters will outline the empirical part of the study. The mono method research consists of a pre-study and a main study to collect the data used in the analyses. The conducted research will be discussed in detail, starting with the presentation of the preliminary study, followed by a more elaborate chapter of the main study. This includes a description of the characteristics of the sample, research design, data collection, measurements of the variables and the analytical strategy.

3. Preliminary study

3.1 Introduction pre-study

As noted previously in the literature, categorization of goods can vary a lot between people. The aim of the pre-study was therefore to determine three goods that by the consumers in the target population are associated with each of the three categories. The products that most strongly are associated with each category will be used in the main study in order to compare the moderating effects of product types on the relationship between maximization and satisfaction post-purchase.

3.2 Methodology pre-study

To determine which category a set of selected products are associated with, an exploratory cross-sectional survey was conducted. The survey was designed with the software program Qualtrics. All questions and instructions were written in English, as respondents were both Dutch

(29)

and international. The survey starts with an introductory text followed by instructions on how to fill in the survey. Definitions of physical products, experiences and credence goods were given prior to the questionnaire, so that every participant would have the same prerequisite to give their responses (see definitions in preliminary survey Appendix 1). After they read the definitions they were asked to place 12 different goods into the category they saw as most fitting. The goods that were expected to be evaluated as physical products were jeans, a necklace, a chair and a TV. Goods expected to be evaluated as experiences were AirBnB booking, holiday abroad, Uber ride and meal at a restaurant. Goods expected to be evaluated as credence goods were vitamin supplements, visit to the dentist, phone reparation and education. In addition respondents were asked to fill in whether they had consumed or purchased the good in the past 6 months, to check how familiar and relevant the goods were for respondents. Lastly, respondents filled in basic demographics. The full survey can be seen in Appendix 1.

The population of the study is consumers living in a modern western country. As the population is very large, the survey uses a convenience non-probability sampling technique to collect primary data. All responses are self reported.The survey was distributed online via social media and direct messages such as email, facebook messenger, facebook groups and Whatsapp. To increase response rate reminders were sent out via private messages. The actual response rate is difficult to measure due to the selected distribution method. The data collection period was limited to 18th- 28th October 2017, which was sufficient to reach the desired minimum amount of responses. Limitations with selected methodology will be discussed in chapters 4.3 and 7.

(30)

3.3 Results Preliminary Study 3.3.1 Sample Description

In the data collection period for the preliminary study a total of 68 responses were collected, out of which 42 were completed (completion rate 61.8%). A general minimum accepted level of respondents for statistically valid data is _{n = 30, thus the sample size for the pre-test is} acceptable. The respondents in the sample age from 18-60 (mean age 31), out of which 59% are female.

3.3.2 Categorization of goods

The purpose of this study was to identify a physical product, an experience and a credence good to be used in the main study that would represent each category. The results of how respondents categorized the selected goods can be seen below in Table 1.

(31)

Table 1_{. Results preliminary study} Physical Product

Experience Credence Good

Jeans 88 % 8.0 % 4.0 % TV 27.9% 62.8 % 9.3% Chair 81.4% 14.0% 4.6% Necklace 86.0% 7.0% 7.0% Holiday 0.0% 97.7% 2.3% Uber 7.0% 60.5% 32.5% Meal at restaurant 9.3% 83.7% 7.0% Education 2.33% 34.9% 62.8% Phone reparation 25.6% 11.6% 62.8% Vitamin Supplement 38.3% 8.51% 53.19% Visit to Dentist 7.0% 14.0% 79.0% Airbnb booking 2.3% 88.6% 9.1%

Highlighted in Table 1 are the goods that a majority of respondents agreed on when choosing an appropriate category (e.g. 88% of the respondents categorized jeans as a physical product). The following three goods were therefore chosen to be used in the main study; _{jeans as a physical} product, a holiday abroad as an experience, and a _{visit to the dentist} _{for the category of credence} goods.

(32)

4. Method main study

4.1 Research design

A quantitative and correlational research is conducted to collect primary data on the relationship between maximizing and post-purchase satisfaction and the moderating effects of type of good and online reviews. The main study consists of an online survey that is pilot tested prior to actual data collection. The survey design is cross-sectional at an individual level with self-reported data by the respondents. The survey was designed and administered in English with the online software program Qualtrics.

The survey consisted of an introductory text, general instructions and the actual set of questions. Anonymity was guaranteed, and during the survey the ‘Force response’ function was activated to avoid missing data within responses. The questions in the survey were divided into four main parts. Most questions asked respondents to rate their response on a 5-point Likert scale ranging from ‘_{strongly agree’ to ‘strongly disagree’. In the first part respondents were asked to} respond to questions related to the maximization trait. In the second part the cognitive dissonance post-purchase was measured. Respondents were asked to think of the last time they purchased and/or consumed a the good (jeans, holiday abroad, reparation of phone), after which questions related to post-purchase dissonance, the importance of the purchase and the impact of online reviews were posed for each of the three goods. The third part asked respondents to rate their tendency to write and consult online reviews in general. Lastly, respondents were asked to fill in a few general questions about demographics. The survey can be found in Appendix 2.

(33)

4.2 Measures

The following part describes the measures used for each variable. To ensure construct validity, most of the scales used in the survey are previously validated and adopted as little as possible.

4.2.1 Maximization

For measuring the independent variable (IV) ‘tendency to maximize’ two separate scales are used as proposed by Cheek & Schwartz (2016).

To measure the ‘_{goal of choosing the best’ the validated Seven Item Maximizing}

Tendency Scale (MTS-7) by Dalal _{et al. (2015) is used. All items are measured on a 5-point} Likert scale(1= strongly agree, 5= strongly disagree). The MTS-7 is a newly improved version of the MTS scale by Diab _{et al.}_{(2008) with a Cronbach's Alpha of 0.82 in the original study. In} the current study the reliability was computed to ⍺= .765, with no counter-indicative items. For the scale items see Appendix 3.

To measure ‘_{maximization strategy’ the the alternative search subscale of the MS-S} (Nenkov _{et al., 2008) is used. This two-item scale is rated on a 5-point Likert scale (} 1= strongly agree, 5= strongly disagree). Items can be seen in Appendix 3. The construct validity of this subscale is a little lower than a usual accepted level of Cronbach's Alpha = 0.62, however the subscale was included in the survey due to the recommendation of Cheek & Schwartz (2016). In the current study the Cronbach’s Alpha was found to be very low, 𝛼 = .255. This can partially be argued to be due to the low number of items. Because of the low Cronbach’s Alpha and low inter-item correlation the scale was not included in the analysis, a decision further supported by the factor analysis that is described later.

(34)

This leaves only the MTS-7 scale by Dalal et al. (2015) to measure maximization amongst respondents in the analysis. For the hypotheses testing the researcher was interested in looking at the relationship between high maximizing tendencies and low satisfaction, but no specific level was chosen to determine ‘high’ or ‘low’ tendency to maximize.

4.2.2 Post-Purchase Cognitive Dissonance

To determine the dependent variable (DV) ‘satisfaction’ the participants’ cognitive dissonance post-purchase was measured with the validated 22-item scale by Sweeney _{et al. (2000). All items} are measured on a 5-point Likert scale ( 1= strongly agree, 5= strongly disagree). The scale is widely used and has an acceptable Cronbach’s Alpha of 0.71 which is why the current study chose to use this despite the many items. Current study measured a very high Cronbach’s 𝛼 = 0.947. The scale was used three times in every survey to evaluate post purchase dissonance for each good. Items can be seen in Appendix 3.

4.2.3 Type of Good

The first moderator, type of good, is used in the survey when testing the cognitive dissonance post-purchase for each good. The satisfaction post purchase is thus measured separately for physical product, experience and credence good. The goods used in the analysis are determined in the preliminary study and further altered after the pilot study (see chapter 3 and 4.4). The final selection of goods are _{Jeans, Holiday abroad and Reparation of Phone.}

(35)

4.2.4 Online Reviews

The impact of online reviews on the decision is measured with three items that are not from a previously validated scale. Items in the scale are ‘ _{I consulted online reviews before making the}

purchase decision_{’, ‘The online review(s) had an impact on my decision’ and ‘The online} review(s) made my decision easier_{’. The items were measured on a 5 point Likert scale (} 1= strongly agree, 5= strongly disagree).In the current study the Cronbach’s Alpha of the 3 items is 0.965, ensuring reliability of the measurement. The variable was used in the study to determine if there was an online review present in the decision making process and whether that had an impact on the decision.

4.3 Sampling and data collection

The population of the study is all consumers in the western countries. As the sampling frame is very large, a non-probability convenience sampling technique is chosen for the appropriate data collection method. The survey was distributed online via email and social media (Instagram, Facebook and Whatsapp). An advantage of online surveys is the relatively easy administration, low cost and large reach, which is why this method was selected. A disadvantage with online reviews however is the much lower response rate in general when compared to paper surveys. Prior studies show that online surveys have a 33% response rate compared to 56% for paper surveys (Baruch, 1999; Nulty, 2008).

First an URL link to the survey was posted in various Facebook groups with a brief introductory message. The link was further posted in a video in the researchers Instagram account, asking followers to respond. In addition, to increase number of responses, direct

(36)

personal messages were sent via Messenger, Whatsapp and email. To increase response rate a reminder to complete the survey was sent out a few days after the initial messages. The survey was moreover made mobile-friendly to make it as convenient as possible for respondents. The response rate of this study is difficult to calculate, as there is no data available on how many people have been reached and exposed to the link online. The response rate however is expected to be in line with the average 33% of online surveys.

To increase reliability of data a check for careless responses was included in the form of a bogus item (Meade and Craig, 2012), namely “ _{I have visited all countries in the whole world} (please respond with only strongly disagree to this question) _{”. Participants answering incorrectly} to this item are considered careless respondents and thus deleted.The data collection period was limited to the 18th November to 17th December 2017. All responses were voluntary and the reported data is strictly confidential.

4.4 Analytical strategy

To check for possible logistical problems and the overall feasibility of the survey, a pilot test was performed prior the actual data collection period (Teijlingen and Hundley, 2002). The main survey in this research was pilot tested on 4 people, and based on the results following points were noted. Even though the largest percentage of respondents in the preliminary study (79%) placed a dentist appointment as a credence good, another one was selected for the main survey. A respondent pointed out how deciding on the purchase of a dentist appointment often does not require an active decision making moment, because most dentist regularly invite their patients for a check up. Although the previous literature considers medical appointment as a classic

(37)

example of a credence good, it is not relevant in this particular research when looking at the active decision making of consumers. This is why the credence good ‘reparation of phone’ replaced the dentist appointment, as it is seen as a credence good by the second largest percentage of participants in the pre-study.

To analyze the data and to test the hypotheses of the study regarding the relationship between maximizing and cognitive dissonance post purchase, together with the moderating effects of product type and online reviews, the online statistical software program SPSS 24 is used. The data set is cleaned for errors, missing values and careless responses. After this, preliminary analyses are made. The descriptive statistics are depicted, after which reliability of scales and a principal factor analysis is performed. This is followed by the computing of scale means and examining the Pearson Correlation. When testing the the hypothesized relationships the SPSS macro PROCESS by Hayes (2012) is used and run for each type of good, in total three times (Process model 2, see Figure 3).

4.5 Preparing the data

Prior to the hypotheses testing following preparatory steps are performed. The raw dataset was downloaded from Qualtrics to SPSS 24. Missing data was checked for, and 124 responses with missing values were excluded listwise together with the 19 respondents answering incorrectly to the bogus item. Data was restructured from variables to cases due to the large amount of responses on the CD scale for three types of goods. Birth year was computed into age and gender was recoded into a dichotomous variable so that it could be included in the analysis. One participant that did not want to specify the gender was grouped with male respondents .Further,

(38)

the 3 different types of goods were re-coded into dummy variables. No counter indicative items were included in the survey, however the results from the Likert scales when measuring CD was re-coded so that a lower score indicated less satisfaction. This was done purely for logistical reasons as originally a high score of cognitive dissonance would indicate low satisfaction. A similar re-coding was done for the impact of online reviews, so that a low score indicated that they did not consult online reviews and OnRev did not have an impact on their decision. A frequencies check for each variable was performed. The data was checked for minimum and maximum values as well as number of valid and missing cases, all indicating that the data was suitable for analyses.

5. Results main study

In the following part the results of the data analysis will be presented. First, a the sample is described, after which preliminary analyses are reported. Finally the main results of the hypotheses testing are presented.

5.1 Sample Description

In the data collection period a total of 359 respondents were collected. After the survey was closed, 124 incomplete responses were deleted. The 19 respondents answering incorrect to the bogus item were further filtered out.This created a final sample of _{n = 216 responses that were} included in the analysis. The completion rate of the survey is calculated to be 60. 2%. The sample consists of participants with an age varying from 19-67 years and an average age of 26

(39)

(SD = 7.47), out of which 73% were female and 27% male. The demographics can be seen in the Appendix 4.

5.2. Preliminary analyses

A Principal Component Analysis (PCA) was conducted to explore the interrelationship amongst variables. Prior to the PCA a few conditions were checked to ensure that the analysis could be performed. The Kaiser-Meyer-Olkin (KMO) measure established the sampling adequacy ( >.6) for the analysis with a value of KMO = .910. A high KMO value indicates low proportion of variance among variables that might be common variance (Kaiser, 1974). In addition, when looking at the Bartlett’s test of sphericity (Bartlett, 1954) statistical significance was reached χ² (496) = 16388.202, p < 0.001, indicating that correlations are sufficiently large between items for the analysis.

To obtain eigenvalues for each component a preliminary analysis was run, which suggested an eight factor solution explaining 72.56% of the total variance. When examining the scree plot (Pallant, 2011) the graph suggested a levelling off at the 8th factor. Therefore, eight factors were extracted and rotated with Varimax, Kaiser Normalization. The component matrix table showed few loadings on factors 6-8, with some cross-loadings, implying that fewer amount of factors would be more appropriate. Models with 7, 6, 5 and 4 factors were run before deciding on a five factor model as the most appropriate. Noted in the process was that the two items of MS-S scale did not seem to belong in a lower than 8 factor-loading solution, and were thus deleted. The five-component solution explained a total of 65.86% of the variance, which is considered sufficient (Pallant, 2011). Further, there are high loadings on each item and no

(40)

cross-loadings. Items that cluster on the same factors suggest that factor 3 represents maximizing, factor 4 represents online reviews and 1, 2 and 5 represents different aspects of cognitive dissonance. This is in line with expectations, as Sweeney _{et al. (2000) describes three} separate aspects in the scale; emotional, wisdom of purchase and concern over deal. Appendix 6 shows the factor loadings after rotation.

Next, scale reliability was computed for all variables. The measuring of the impact of online reviews did not adapt a previously validated scale, but was a variable developed for this study. The rest of scales used are previously validated scales with a Cronbach’s Alpha higher than 0.7 in their original paper. In the current study all scales used have a high reliability thet could not be heightened by deleting any items (see Table 2). As previously noted, the maximization subscale by Nenkov et al. (2008) is not included due to the low reliability 𝛼 = 0.26. This might partially be explained by the few items ( _{n = 2), and further when checking the} inter-item correlation the value of .148 is below recommendations (Briggs and Cheek,1986). An overview of the scales used in this study can be seen in Appendix 3. Scale means can be seen below in table 2.

(41)

Lastly the normality of the data was measured. Shapiro-Wilk test for normality indicates that all variables have a significance below .05, thus variables are not normally distributed. Scores for skewness and kurtosis are not optimal, however they land within the acceptable range of ∓ 2 (George & Mallery, 2010). In addition, as the current sample size is large (n > 200) the skewness and kurtosis values do not cause consequential differences (Tabachnick, Fidell, & Osterlind, 2001). A high negative skewness is found for cognitive dissonance (-1.54), indicating that a large group of participants scored low on the CD scale, and thus were satisfied with their purchase. The limitation this puts on the study is discussed later. Maximizing is fairly normally distributed, confirmed by the bell-shaped histogram, with only a slight skewness to the left (0.24). The impact of online reviews has a slight skewness to the right (-0.22), but with a high peak for score 1 (highly disagree) indicating that many respondents consulted and were affected by online reviews. As the distribution is not normal, the scale is recoded into two groups of ‘ _{low’ (scores} 1-2.99) and ‘_{high’ (scores 3-5). The ‘low’ group were people that did not in general read or get} influenced by online reviews prior to their purchase. The two groups were almost equally large, with 52.8% in the low group and 47.2% in the high. See values for skewness and kurtosis in Appendix 5.

5.3 Correlation analysis

A correlation analysis is run to investigate the strength and direction of relationships between variables. The results are presented below in Table 3.

(42)

Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 1. SAT 4.42 .68 1 2. MAX 2.51 .67 .04 1 3. OnRev 0.47 .50 -.11** -.06 1 4. Jeans 0.33 .47 .02 - -.41** 1 5. Holiday 0.33 .47 .19** - .34** - 1 6. Reparation 0.33 .47 -.21** - .07 - - 1 7. Gender 0.73 .44 −.08* -.14** .02 - - - 1 8. Age 26.34 7.46 .06 .15** .06 - - - -.12** 1 9. RW 1 1.44 .50 -.11** .06 -.09* - - - .03 -.10** 1 10. RW2 3.51 1.19 -.03 .05 -.12** - - - −.06 -.20** .43** 1 11. Time 3.56 1.12 -.12** .02 .05 -.31** -.28** .59** .05 -.02 -0.3 .00 1 12. Importance 3.56 1.12 -.03 -.12** .18** -.17** .281** -.11** -.12** -.01 -.02 -.07 -.28** 1

** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed) - Not applicable

Included in the correlation matrix is the independent variable of tendency to maximize (MAX), dependent variable of satisfaction post-purchase (SAT), the first moderator ‘type of good’ (Jeans, Holiday, Reparation) and second moderator ‘online reviews’ (OnRev) together with control variables gender, age, average importance of purchase, average time of purchase, previous writing of online reviews ( RW1) and likeliness of writing an online review (RW2). Variables SAT and OnRev were measured three times separately in the survey for each type of good, but the rest of the variables were measured only once per respondent. This explains the gaps in the correlation matrix (marked with -). Effect sizes are interpreted in line with suggestions of Cohen (1988). Next, most relevant correlations are discussed.

Can spending money make some of us happier? : the role of maximization in post-purchase satisfaction for three types of goods and the impact of online reviews