“What review shall I read?” How
consumer purchase decision phases and
online review useful votes affect selective
review processing.
Student: Marije J. N. Kniep (Annemarijn), Ba Student number: 10445420
Date: 30-06-2017
Supervisor: Dr. M. L. Fransen
Research Master’s Thesis Communication Science Graduate School of Communication
2 ABSTRACT
Consumers frequently rely on online product reviews when making their purchase decisions. However, the large number of available reviews makes it impossible and also undesirable to read all of them. This is why consumers often only select a small amount of reviews to process. In this study it is expected that this selection and processing of (positive and
negative) online reviews depends on the purchase decision phase of the consumer. Therefore, this paper examines the selective processing of online reviews before (pre-) and after (post-) a purchase decision has been made. Based on Festinger’s theory of cognitive dissonance
(1957), no selection bias was expected for people in the pre-purchase decision phase, whereas a preference for decision-confirming information was expected for people in the
post-purchase decision phase. Furthermore, since decisions are known to be influenced by
someone’s social environment, it was expected that the amount of review useful votes would influence which reviews would be selected. The results of two experiments indicate that decision confidence is an important factor that affects the selection and perception of reviews. Nevertheless, consumers select in general more negative than positive review titles.
Furthermore, with regard to the perception of reviews,it seems that instead of having a confirmation bias in the post-purchase decision phase, people rather have a negativity bias in the pre-purchase decision phase. Review useful votes did not seem to have an effect on the selection of review titles. Insights of this research can help managers and researchers to gain a better understanding of consumers’ online review processing.
3 Introduction
Online product reviews, a prominent form of electronic Word Of Mouth (eWOM), are
perceived as one of the most powerful persuasive tools of the 21st century (King, Racherla, & Bush, 2014; Mafael, Gottschalk, & Kreis, 2016). In fact, 72% of the Dutch online consumers indicate that online reviews have an impact on their purchase behaviour (dtg, 2015). It is therefore not surprising that online reviews appear almost everywhere online nowadays: on search engine result pages, on company websites, and on review sites. A consequence of the large number of consumer opinions available is that it becomes impossible and also
undesirable to read all reviews (Park & Lee, 2008). Results from the Local Consumer Review Survey showed that 90% of the people read less than 10 reviews before deciding whether to purchase something from a business or not (BrightLocal, 2016). Thus, people base their final purchase decision on a relatively small selection of all available online product reviews.
The selection of online product reviews depends among others on someone’s purchase decision phase. Someone who has not made a purchase decision yet (pre-purchase decision phase) is expected to search for both positive and negative information to make a balanced purchase decision (Fischer & Greitemeyer, 2010). However, someone who has already made a purchase decision (post-purchase decision phase) might not even be looking for reviews, but will still be confronted with them (Mafael, et. al., 2016). Imagine someone notices a nice restaurant on a good location filled with customers. At that moment he or she decides to have dinner there someday. However, when searching for the restaurant online to make a
reservation, one will most likely be confronted with positive as well as negative reviews. Either on the search engine result page, on the restaurant’s website, or both. In this situation, this person might perceive positive reviews as more relevant and persuasive than negative reviews to confirm one made the right decision. The preference for information that is in line with one’s perceptions, attitudes, and behaviour is called a confirmation bias (Jonas,
Schulz-4
Hardt, Frey, & Thelen, 2001), which may result in selective exposure to decision-confirming information (e.g. Fischer, Greitemeyer, & Frey, 2008).
Besides the cognitive factor of someone’s purchase decision phase, also external review characteristics like review ‘useful’ votes affect the selection of reviews (Willemsen, Neijens, Bronner, & De Ridder, 2011). Many webshops and review websites include peer-rating systems to help customers deal with the large amount of online consumer reviews. These systems enable consumers to vote if they consider a review useful in their purchase decision-making (Willemsen, et. al., 2011). The votes are often used as cues to select reviews more easily (Gottschalk, & Mafael, 2017). It is therefore expected that the tendency to select only positive reviews in the post-purchase decision phase will be disrupted when negative reviews have more useful votes than positive reviews. This will also disrupt the balanced selection of review titles in the pre-purchase decision phase, because of the strong impact of social norms on individual behaviour (Cialdini, Reno, & Kallgren, 1990). So, the main research question is formulated as follows: To what extent does the consumer decision-making phase have an effect on the selection and the perceived relevance and persuasiveness of online reviews and is this effect moderated by the proportion of review useful votes?
The role of individual characteristics that influence consumers’ perception and processing of eWOM information continues to be a relevant gap in eWOM research (King, Racherla, & Bush, 2014). This study contributes to the literature by examining whether different consumer decision-making phases lead to a different review selection and
perception. Also, it combines research on consumer characteristics with research on online review cues by examining the disrupting effect of useful votes on selective exposure. By combining the impact of consumer characteristics and review information cues on eWOM selection, managers of webshops and review websites can use the results to present online reviews in a more useful way for their visitors.
5 Selective exposure as a form of confirmation bias
The “pull” nature of the Internet environment provides people with the opportunity to select and process whatever information they want while ignoring other information (Bimber, & Davis, 2003). Usually, people prefer information that is in line with their decisions or
opinions on information that contradicts their decisions or opinions. This is referred to as the confirmation bias (Jonas et. al., 2001). Because of the confirmation bias, information that is ignored usually concerns information that is incongruent with one’s decision or opinion (Festinger, 1957). The bias in information preference thus logically results in a bias in the selection of information to process. This phenomenon is called selective exposure (e.g.
Fischer, Fischer, Weisweiler, & Frey, 2010; Fransen, Smit, & Verlegh, 2015). The underlying psychological motivation for selective exposure is often explained in the context of cognitive dissonance theory as a strategy to avoid or reduce post-decisional conflicts (Festinger, 1957; Frey, 1986; Tanford, & Montgomery, 2015; Liang, 2016).
In 1957, Festinger proposed his theory of cognitive dissonance. Cognitions are elements of knowledge that people have about their behaviour, attitudes, and environments. According to cognitive dissonance theory, two cognitions can either be related or unrelated to each other. If they are related, they can be either consonant or dissonant. Two cognitions are consonant when one cognition supports the other and they are dissonant when one cognition counteracts the other (Festinger, 1957). For example, two consonant cognitions are “That is a beautiful restaurant” and “The food quality of that restaurant is good”. Two dissonant
cognitions are “The food quality of that restaurant is good” and “That restaurant serves frugal portions”.
Cognitive dissonance theory builds upon the idea that people strive for cognitive consonance; have the tendency to hold their attitudes and beliefs in harmony. Conversely, cognitive dissonance is a state of mental imbalance or disharmony that people try to avoid or
6
reduce. Usually they do this by actively avoiding information from which they expect
dissonance to occur or by paying more attention to decision-consistent information (Festinger, 1957; Frey, 1986; Fransen, et. al., 2015). For example, Brock and Balloun (1967) found that smokers paid more attention to a message stating that smoking is no serious health risk than to a message stating that smoking is a serious health risk. For non-smokers the opposite pattern was found. Also in more recent studies conducted in the context of decision-making, results indicated that people have the tendency to prefer decision-consistent information (Fischer, Schulz-Hardt, & Frey, 2008; Fischer, et. al., 2010; Jonas, et. al., 2001). For example, Fischer, Schulz-Hardt, and Frey (2008) asked participants to decide whether the contract of a store manager should be extended or not. They received information to base a preliminary decision on. Subsequently, they were allowed to select additional information for their final decision. With ten extra pieces of information available, participants selected much more information in line with their preliminary decision (over 80%) than information contradicting their
preliminary decision. In the case of online review selection it can therefore be expected that consumers in the post-purchase decision phase have the tendency to select positive reviews rather than negative reviews. This is because they are already in favour of a product or service and want to confirm that they made the right decision. Negative reviews about the product or service will be avoided as much as possible, as these contain shortcomings of the product of choice. These reviews would confront the consumers with negative aspects of their decision, causing cognitive dissonance.
The selective exposure behaviour of consumers in the post-purchase decision phase as described above is expected to deviate from consumers in the pre-purchase decision phase. This is because people in the pre-purchase decision phase have not committed themselves to a product or service yet. According to Festinger (1957): “... the preaction or predecision
7
(p. 126). Therefore, it is expected that people in the pre-purchase decision phase will not have the tendency to avoid or prefer certain information. Their main focus is to make an accurate decision using the available reviews. This would thus lead to a balanced selection of positive as well as negative review titles. This leads to the following hypotheses:
H1: Consumers in the post-purchase decision phase have a confirmation bias such that they
select more positive than negative review titles while consumers in the pre-purchase decision phase select an equal amount of positive and negative review titles.
H2: Consumers in the post-purchase decision phase have a confirmation bias such that they
select more positive and less negative reviews than consumers in the pre-purchase decision phase.
Perceived review relevance and persuasiveness in different purchase-decision phases
Since it is expected that consumers in different purchase decision phases differ in their selection of positive and negative reviews, it might be expected that they also differ in their perception of these reviews. This is because consumers in the post-purchase decision phase are trying to reduce the experienced cognitive dissonance when they are confronted with a negative review. On the other hand, consumers in the pre-purchase decision phase cannot experience cognitive dissonance since they have not committed themselves to a product yet. They have the goal to make the best product decision. Therefore, it is expected that consumers in the pre-purchase decision phase perceive both positive and negative reviews as relevant for their final purchase decision (Festinger, 1957). A combination of both positive and negative reviews is needed to be able to make a balanced choice. Furthermore, in research on
persuasiveness and online reviews it was found that as long as a set of reviews displays conclusive information and contains high quality arguments, this information will be
8
perceived as persuasive. The valence of the reviews did not influence this outcome (Mudambi, & Schuff, 2010).
Contrary to people in the pre-purchase decision phase, people in the post-purchase decision phase have committed themselves to a product or service. Their goal is to confirm they made a good product decision. Therefore, it is expected they will not only show a bias in their selection of reviews, but also in their perception of reviews. In previous research, Mafael et. al. (2016) found that consumers perceive positive (negative) arguments in online reviews as more (less) persuasive when having a positive attitude towards a brand. This makes it interesting to examine whether a purchase decision (not based on a brand) could cause the same effect. When people have made a purchase decision, it is also expected this would result in minimizing the relevance of disconfirming (negative) information, to reduce the negative impact. The conclusiveness and argument quality of the information should not influence this outcome (Feather, 1963; Mafael, et. al., 2016).Thus:
H3: Consumers in the post-purchase decision phase perceive negative reviews as less relevant
and less persuasive than positive reviews while consumers in the pre-purchase decision phase perceive negative and positive reviews as equally relevant and persuasive.
H4: Consumers in the post-purchase decision phase perceive negative reviews as less relevant
and less persuasive than consumers in the pre-purchase decision.
The disrupting effect of review useful votes
Until now the focus has been on the influence of people’s cognitive stage on the selection and perception of positive and negative online reviews. However, decisions are rarely made in complete isolation. Instead, they are influenced by someone’s social environment (Cialdini, Reno, & Kallgren, 1990). An important review property that serves as social environment in
9
selecting online reviews is the amount of review useful votes (Gottschalk, & Mafael, 2017). In a recent study on selective e-WOM processing, Gottschalk and Mafael (2017) found through interviews and a survey that review useful votes are often used as information cues to consumers to select reviews more easily. This is reasonable considering that people take the expectations and behaviour of others into account when they decide what individual actions are appropriate (Schultz, et. al., 2007). The expectations and behaviours of others are defined as social norms, and they strongly influence someone’s preferences and behaviour.
In the literature, social norms are subdivided in descriptive and injunctive norms (Cialdini et. al., 1990). Descriptive norms provide evidence of what will presumably be an effective and adaptive action to undertake (Jacobson, Mortensen, & Cialdini, 2011; White, & Simpson, 2013). They describe what most other people do, whereas injunctive norms describe what most others approve or disapprove of (Cialdini, et. al., 1990). Review useful votes can be categorized as descriptive norms because they show which reviews were most utilized by other consumers. For example, when an online review has 98 review useful votes, it says that 98 other people used this review to base their final purchase decision on. This provides evidence that this review will contain valuable information for someone who is about to purchase a product.
Since a high amount of review useful votes serves as evidence that a review contains valuable information, it will become more difficult to ignore or refute these reviews compared to reviews without or with a low amount of review useful votes. In previous studies it has already been found that strong evidence makes it difficult to refute counter attitudinal information (e.g. Ahluwalia, 2000). It decreases avoidance and couterarguing of this
information. In the case of online reviews, it can for this reason be expected that consumers will always have the tendency to select the reviews with the highest amount of review useful votes, independent of their purchase decision phase. They cannot ignore the evidence that
10
these reviews were used by many other consumers. Therefore, a high amount of review useful votes for decision-incongruent (i.e. negative) reviews is expected to disrupt the confirmation bias of someone in the post-purchase decision phase. More negative reviews will be selected. It will also disrupt the balanced selection and perception of positive and negative reviews of consumers in the pre-purchase decision phase. Just as consumers in the post-purchase decision phase, they will have the tendency to select the reviews with the highest amount of review useful votes, independent whether these reviews are positive or negative. Thus:
H5: Consumers in any purchase decision phase select more reviews with a large amount of
review useful votes rather than with a small amount of review useful votes.
H6: The effect of purchase decision phase on the selection of reviews is moderated by review
useful votes such that the effects disappear when reviews have a high amount of review useful votes.
Experiment 1: Selection, perceived relevance, and persuasiveness of reviews.
The first experiment focuses on the influence of people’s cognitive stage on selective review processing. Therefore, this experiment tests hypotheses 1-4, concerning the impact of the purchase decision phase on the selection and the perceived relevance and persuasiveness of positive and negative reviews. A scenario was used wherein people had to choose between different products, had to select review titles and had to evaluate a positive and a negative review on their relevance and persuasiveness.
Research design and participants
Hypotheses 1-4 were tested using a 2 (purchase decision phase: pre vs. post) x 2 (review valence: positive vs. negative) mixed factorial design with purchase decision phase as a
11
between subjects variable and review valence as a within subjects variable. Participants were personally invited or approached through e-mail or social media to take part in the study. A link was shared, which led the participants to the experiment. Participation was voluntarily and among the participants a gift voucher of €15 was allotted.
Of the 92 people that started the questionnaire, six dropped out somewhere during the study. Also, one person did not follow the instructions correctly. Of the six people that dropped out, two completed the questionnaire for more than 50%. These were kept for analysis. The results of the remaining dropped out participants and of the person who did not follow the instructions were excluded from the analyses using listwise deletion. The
remaining sample consisted of 87 participants. A larger percentage was female (60.9%, n = 53) than male (36.8%, n = 32). Two participants did not disclose their gender. The ages of the participants ranged from 18 to 67 and the mean age was Mage = 41.15 (SD = 14.31). Almost half of the sample indicated having completed a moderate high level of education (HBO) (47.1%, n = 41). 23% completed a higher level of education (University) (n = 20). 25.2% indicated having completed a lower level of education (high school, MBO) (n = 21) and two participants completed a different kind of education (2.3 %). Two participants did not disclose their education level.
Procedure
Participants were randomly assigned to one of the two between-subjects experimental conditions: To the pre-purchase decision condition 43 people were assigned and to the post-purchase decision condition 44 people were assigned.
After agreeing to the informed consent, participants were directed to the manipulation of the purchase decision phase. They read a scenario and then continued to a page containing three products including product descriptions and important specifications. They were
12
instructed to read the information carefully and were able to click through to continue the experiment after one minute. In this part, participants in the post-purchase decision condition were asked to make a purchase decision and continued to decision confidence measures. Participants in the pre-purchase decision phase only answered filler questions about the amount of available information. Participants in both conditions subsequently continued to twelve product review titles. They were instructed to select five titles they thought were interesting to read. After their selection, participants were exposed to a negative and a positive review in a random order. Per review, they answered questions regarding perceived review relevance and persuasiveness. Then they answered a final purchase decision question and control questions on decision involvement and general use of online reviews. As afinal part of the experiment, participants answered demographic questions (gender, age, and education level), had the option to share remarks and were invited to take part in the lottery. Also, they were given the opportunity to leave their e-mail address if they were interested in receiving more information about the research and the results. Participants were thanked for their participation and asked to submit the questionnaire. The experiment took 10-15 minutes.
Research materials of the independent variable
For the manipulation of the purchase decision phase, participants read a scenario on choosing an alarm clock radio as a birthday present for a good friend. It was the participant’s task to decide which of the three alarm clock radios was the best one to give. In the pre-purchase decision phase, participants could use the product description, specifications and reviews to make a final decision, while participants in the post-purchase decision phase were asked to make a purchase decision immediately after exposure to the product description and
13
before exposure to review titles and reviews, while participants in the post-purchase condition did choose a product before this exposure. The scenario is included in Appendix A.
An alarm clock radio was used as birthday present for several reasons: First, it is a flexible product, with different and unique shapes, sizes, specifications, and prices. Still, different alarm clock radios are comparable since they all have the same function. Second, the alarm clock radios people could choose were all higher-priced. A higher price in combination with unique features causes a higher attached value and product involvement in choosing a product (Martin, 1998). Finally, the differences in design and specifications enables
participants to develop a strong preference for one of the products (Martin, 1998). Participants had to select an alarm clock radio for someone else to make sure they never had to choose a product they would not want in reality. For example, when someone always uses his or her Smartphone as an alarm clock, this person would not at all be interested in buying an alarm clock radio. Therefore, a good friend was chosen as the receiver of the product. This solves the personal interests problem, while the involvement is expected to be good since someone will be motivated to choose something nice for a good friend.
The three products, product descriptions, and specifications were taken from mediamarkt.nl and slightly modified. To make sure the brand would not influence the outcomes, the brand name was removed from the product pictures and descriptions. One of the product descriptions was shortened to keep all descriptions the same length. Also, not all the same kind of specifications were available for all alarm clock radios. Therefore, other Internet websites were used to search for and add the missing information in the experiment material. The product information is included in Appendix B.
As described in the procedure, participants had to select five out of twelve review titles and then continued to two complete reviews which they had to evaluate. The review titles and reviews were based on existing reviews at google.nl/shopping. The review titles were either
14
directly copied or they were composed from existing review titles in order to meet the requirements of the study by being clearly positive or negative and all different from each other. The review titles participants could choose of are included in Appendix C. The two complete reviews were created by combining (short) existing reviews to meet the
requirements of being conclusive and of high argument quality. The reviews were equal in word count and in number of arguments included. They are included in Appendix D.
Measures
Decision confidence
Participants in the post-purchase condition rated two confidence items after they made a purchase decision. One on their confidence of the product decision and one on the certainty of their choice. The items were rated on a seven-point confidence scale and a seven-point
certainty scale (1 = not at all, 7 = completely). The means of the items were used to estimate a scale on decision certainty (M = 4.99, SD = 1.25, Cronbach's α = .91).
Selection of reviews
The dependent variable of review selection was measured by the actual behavior of the participants. They were instructed to select five out of twelve review titles. Their selection was used to assess whether participants in one purchase decision condition selected more or less positive (M = 2.14, SD = 1.31) and negative (M = 2.86, SD = 1.31) reviews compared to participants in the other purchase decision condition.
Relevance and persuasiveness
After exposure to both complete reviews (i.e. positive and negative), participants rated the dependent variables of perceived relevance and persuasiveness for both reviews. To measure
15
review relevance, a four item, seven-point Likert scale (1 = totally disagree, 7 = totally agree) was used. The items were: “This review is relevant for my situation”, “This review is important for my choice”, “I can use this information”, and “This review is helpful”. The means of the items were used to estimate a scale of perceived relevance of positive reviews (M = 4.86, SD = 1.22, Cronbach's α = .92) and of negative reviews (M = 5.22, SD = 1.32, Cronbach's α = .93).
The dependent variable of perceived persuasiveness was measured with an existing two-item seven-point scale (1 = not at all, 7 = very much). Items included were: “How strong are the arguments presented in this online review?” and “How convincing is this review?” (Munro, Ditto, & Lockhart, 2002). The means of the items were used to estimate a scale of perceived persuasiveness of positive reviews (M = 4.47, SD = 1.30, Cronbach's α = .88) and of negative reviews (M = 4.92, SD = 1.31, Cronbach's α = .94).
Decision involvement
In the experiment, two control variables were included. The first one was decision
involvement, which was included to control how serious participants took part in the scenario. The variable of decision involvement was rated on an existing, three-item scale with the product adjusted for the experiment (Bojanic, & Warnick, 2012). The scale consisted of the following items: “How important was it to make a right choice of this product?” (measured with a seven-point importance scale), “How interested were you to select the best alarm clock radio?” (measured with a seven-point interest scale), and “How concerned were you with choosing the best alarm clock radio?” (measured with a seven-point concern scale). The means of the items were used to estimate a decision involvement scale (M = 5.33, SD = 1.08, Cronbach's α = .82). A higher score indicated a higher decision involvement.
16 Online review importance
The second control variable concerned general online review importance. This variable was included to assess whether participants used and valued reviews in their ordinary life. The variable was measured using a yes/no item and a seven-point importance scale. The yes/no item was: “Do you ever use online reviews before you decide to buy a product?” (yes = 95.3%, n = 81). The scale item used was: “Do you think online reviews are important in making a purchase decision?” (M = 5.35, SD = 1.25).
Results
Randomization checks
For a randomization check of gender across the two purchase decision phases, a crosstab with Chi-square test was performed. Gender was used as dependent variable and the purchase decision phase as independent variable. The analysis showed that the participant’s gender did not significantly differ between the purchase decision phases (χ² (1) = .13, p = .716).
For a randomization check of age, education level and the control variables of decision involvement and online review importance across the two purchase decision phases, One-way ANOVAs were conducted. For the yes/no control variable measuring online review
importance, a crosstab with Fisher’s Exact test was performed since 50% of the cells had an expected count below five. The participant’s individual characteristics of age, education level, decision involvement and online review importance were used as dependent variables and the consumer purchase decision phase was used as independent variable. The analyses showed that neither age (F (1,82) = .19, p = .666), nor education level (F (1,83) = .27, p = .608), nor decision involvement (F (1,83) = .81, p = .371), nor online review importance (F (1,83) = .04, p = .840), or yes/no online review importance (Fishers Exact p = .824) were significantly
17
different between the different purchase decision conditions. Therefore, none of these variables were included as covariates in the further analyses.
Hypotheses testing
All hypotheses were tested using SPSS version 22. The first two hypotheses to be tested concerned the selection of review titles in and between different purchase decision phases. To test the hypotheses, a repeated measures ANOVA was used. The purchase decision phase condition was used as the between-subject variable and the selection of positive and negative review titles was used as the within-subject variable. The ANOVA yielded a significant main effect of selected review titles (F (1, 85) = 6.53, p = .012, η² = 0.71). On average, participants selected more negative review titles (M = 2.86, SD = 1.31) than positive review titles (M = 2.14, SD = 1.31). No interaction effect between the purchase decision phase and the selection of reviews was found (F (1, 85) = .44, p = .510). The purchase decision phase did not have a significant effect on the selection of reviews. Therefore H1, predicting that consumers in the post-purchase decision phase selected more positive than negative review titles while
consumers in the pre-purchase decision phase selected an equal amount of positive and negative reviews is not confirmed. Also H2, predicting that consumers in the post-purchase decision phase selected more positive review titles and less negative review titles than consumers in the pre-purchase decision phase is not confirmed.
H3 and H4 concerned predictions regarding the perceptions of positive and negative reviews in and between the two different purchase decision phases. Two repeated measures ANOVAs were conducted to test these hypotheses. The perceived review relevance and persuasiveness of positive and negative reviews were used as within-subject variables and the purchase decision phase was used as the between-subject variable. The repeated measure ANOVAs yielded a significant main effect of review relevance (F (1, 84) = 4.33, p = .041, η²
18
= .05) and review persuasiveness (F (1, 84) = 5.73, p = .019, η² = .07). Overall, negative reviews were perceived as more relevant (M = 5.22, SD = 1.32) than positive reviews (M = 4.86, SD = 1.22). Also, negative reviews were perceived as more persuasive (M = 4.92, SD = 1.31) than positive reviews (M = 4.47, SD = 1.30). A significant interaction effect was found between the purchase decision phase conditions and the perceived review relevance (F (1,84) = 5.08, p = .027, η² = .06). Also, a marginally significant interaction effect was found between the purchase decision phase conditions and the perceived review persuasiveness (F (1,84) = 2.79, p = .099, η² = .03). In figures 1 and 2, perceived review relevance and persuasiveness of positive and negative reviews per purchase decision phase are visualized. In the post-purchase decision phase, no significant differences were found between the perceived relevance (F (1,84) = .02, p = .910) and persuasiveness (F (1,84) = .26, p = .613) of positive and negative reviews. However, in the pre-purchase decision phase, significant differences were found between the perceived relevance (F (1, 84) = 9.40, p = .003, η² = .10) and perceived persuasiveness (F (1, 84) = 8.35, p = .005, η² = .09) of positive and negative reviews.
Figure 1: perceived relevance per review Figure 2: Perceived persuasiveness per review
As can be seen in figure 1, in the pre-purchase decision phase negative reviews were perceived as more relevant (M = 5.60, SD = 1.25) than positive reviews (M = 4.88, SD =
19
1.22). Figure 2 shows that negative reviews were also perceived as more persuasive (M = 5.19, SD = 1.34) than positive reviews (M = 4.38, SD = 1.32). These outcomes do not confirm H3, predicting that consumers in the post-purchase decision phase perceive negative reviews as less relevant and less persuasive than positive reviews while consumers in the pre-purchase decision phase perceive negative and positive reviews as equally relevant and persuasive. On the contrary, the outcomes show that consumers in the post-purchase decision phase perceive positive and negative reviews as equally relevant and persuasive, while consumers in the pre-purchase decision phase perceive negative reviews as more relevant and more persuasive than positive reviews.
Whereas H3 focussed on differences within decision phase groups, H4 concerned differences between these groups. Pairwise comparisons of the perceived relevance and persuasiveness showed that perceived relevance (F (1, 84) = 7.64, p = .007, η² = .08) and perceived persuasiveness (F (1, 84) = 3.01, p = .086, η² = .04) of negative reviews were respectively significant and marginally significant different between the different purchase decision phase conditions. People in the post-purchase decision phase perceived negative reviews as less relevant (M = 4.84, SD = 1.29) than people in the pre-purchase decision (M = 5.60, SD = 1.25). Also, people in the post-purchase decision phase perceived negative reviews as marginally less persuasive (M = 4.66, SD = 1.24) than people in the pre-purchase decision phase (M = 5.19, SD = 1.34). Therefore, H4, stating that consumers in the post-purchase decision phase perceive negative reviews as less relevant and less persuasive than consumers in the pre-purchase decision is partially confirmed.
Exploration within the post-purchase decision condition group
Since a decision confidence measurement scale was included in the experiment, the selection and perception of online reviews were also explored within the post-purchase decision
20
condition itself. This way, it could be examined whether a higher decision confidence level would lead to another selection and perception of online reviews or not. This could be expected, since a lower decision confidence indicates that someone is not sure whether he or she chose the right product; is not fully committed to this product. On the contrary, decision confident people are fully committed to the chosen product. Therefore, less confident people might be more interested in negative reviews and less interested in positive reviews than people that are more confident about their decision, to be able to make a fair decision on the chosen product. To assess review title selection per decision confidence level, a linear regression analysis was performed. The selection of positive and negative review titles was used as dependent variable and the decision confidence scale as the independent variable. A significant regression equation was found (F (1, 42) = 12.03, p = .001) with R² = .22. A significant positive effect was found between someone’s decision confidence level and the amount of selected positive and negative reviews (β = .472, t = 3.47, p = .001). An increase of one point in decision confidence indicated a .54 increase in the selection of positive reviews. Thus, the more confident someone was, the more positive and the less negative review titles this person selected.
Linear regressions were also conducted to assess review perception per decision confidence level. In the first two linear regressions, decision confidence was used as
independent variable and the perceived relevance of positive and negative reviews were used as dependent variables. The results yielded a significant regression equation for perceived relevance of positive reviews (F (1, 41) = 13.53, p = .001) with R² = .25. A positive significant effect was found between someone’s decision confidence and the perceived
relevance of positive reviews (β = .50, t = 3.68, p = .001). An increase of one point in decision confidence indicated an increased perceived relevance of positive reviews of .54 point.
21
negative reviews (F (1, 41) = 1.32, p = .257). In two other linear regressions, the perceived persuasiveness of positive and negative reviews were used as dependent variables. Decision confidence was still used as independent variable. For perceived persuasiveness of positive reviews, a significant regression equation was found (F (1, 40) = 11.31, p = .002) with R² = .22. A positive effect was found between someone’s decision confidence and the perceived persuasiveness of positive reviews (β = .47, t = 3.36, p = .002). An increase of one point in decision confidence indicated an increased perceived persuasiveness of positive reviews of .47 point. The perceived persuasiveness of negative reviews was not significantly related to someone’s decision confidence (F (1,41) = .22, p = .642).
Conclusion experiment 1
In sum, no support was found for H1, H2, and H3. With regard to H1 no support was found that consumers in the post-purchase decision condition selected more positive than negative review while consumers in the pre-purchase decision phase selected an equal amount of positive and negative reviews. Also, with regard to H2, no support was found that consumers in the post-purchase decision phase selected more positive and less negative review titles than consumers in the pre-purchase decision phase. In fact, all consumers selected more negative than positive review titles. With regard to H3, no support was found that consumers in the post-purchase decision phase perceived negative reviews as less relevant and less persuasive than positive reviews or that consumers in the pre-purchase decision phase perceived negative and positive reviews as equally relevant and persuasive. Instead, participants in the post-purchase decision condition perceived both positive and negative reviews as equally relevant and persuasive. Participants in the pre-purchase decision condition perceived negative reviews significantly more relevant and persuasive than positive reviews. So, even though H3 was not confirmed in this experiment, differences between different purchase decision conditions were
22
found. That is, people in the post-purchase decision perceived negative reviews as less relevant and marginally less persuasive than people in the pre-purchase decision. Thus, someone’s purchase decision phase had an impact on ones perception of reviews. This
partially confirms H4, predicting that consumers in the post-purchase decision phase perceive negative reviews as less relevant and persuasive than consumers in the pre-purchase decision.
When exploring differences between participants within the post-purchase decision condition itself, findings in the expected directions were obtained: decision confident
consumers selected significantly more positive and less negative reviews than consumers who were uncertain about their decision. Also, decision confident people perceived positive
reviews as significant more relevant and persuasive than people lacking decision confidence. Therefore, a second experiment was conducted, wherein the manipulation was adjusted in an attempt to make people more confident about their decision. Also the moderating variable of review useful votes was included in the second experiment.
Experiment 2: Adjusted manipulation and review useful votes.
In the second experiment, all hypotheses (1-6) were tested. This experiment followed the same procedure as the first experiment. Again, people had to choose between different products, had to select review titles and had to evaluate a positive and a negative review on their relevance and persuasiveness. Only in this experiment, participants did not find
themselves in an hypothetical, but in an actual decision making situation. Also review useful votes were included as moderator variable.
Research design and participants
The hypotheses were tested using a 2 (purchase decision phase: pre vs. post) x 2 (review useful votes: present vs. absent) x 2 (review valence: positive vs. negative) mixed factorial
23
design with purchase decision phase and review useful votes as between subjects variables and review valence as a within subjects variable. Participants were invited through e-mail or (personally) approached through social media to participate in the study. A link was shared, which led the participants to the experiment. Participation was voluntarily and among the participants four Bluetooth speakers were allotted, each with a value of approximately €50.
For the second experiment, a student sample was used. Of the 194 students who started the questionnaire, 29 dropped out somewhere during the study. Of these 29 people, 12 completed the questionnaire for more than 50%. These were kept for analysis. The results of the remaining dropped out participants were excluded from the analyses using listwise deletion. The final sample consisted of 177 participants. A larger percentage was female (62.1%, n = 110) than male (31.1%, n = 55). Twelve participants did not disclose their gender. The ages ranged from 19 to 33 and the mean age was Mage = 22.85 (SD = 2.31). Almost half of the sample indicated having a high level of education (University bachelor/master) (44.6%, n = 79), 36.2% had a lower level of education (high school, MBO) (n = 64), 25.2% indicated having a moderate high level of education (HBO) (n = 21), and one participant had a different kind of education (0.6 %). Twelve participants did not disclose their education level.
Procedure
Participants were randomly assigned to one of the four between-subjects experimental conditions: To the pre-purchase decision x useful votes absence condition 43 people were assigned, to the pre-purchase decision x useful votes presence 43 people were assigned, to the purchase decision x useful votes absence 44 people were assigned and to the
post-purchase decision x useful votes presence condition 47 people were assigned.
As stated in the introduction of this experiment, the procedure was the same as in the first experiment, apart from the actual product decision situation and the review useful votes.
24
Participants found themselves in a realistic product decision situation, since they could win the product they chose. After participants agreed on the informed consent, a page containing two products including product descriptions and specifications was shown to them. When they had read the information, participants in the post-purchase decision condition had to make a product decision and answered decision confidence measures. Participants in the pre-purchase decision phase only answered filler questions about the amount of available
information. Participants in both conditions subsequently continued to twelve product review titles for both products. Per product, they were instructed to select five titles they thought were interesting to read. In two experimental conditions, the review titles were accompanied with review useful votes. After their selection, participants were exposed to a negative and a positive review of both products, in a random order. It was necessary to expose participants to review titles and reviews of both the products to make it possible to compare the products and to choose one at the end of the experiment. Per complete review, participants answered questions regarding perceived review relevance and persuasiveness. Then they answered a final product decision question and control questions on decision involvement and general use of online reviews. As afinal part of the experiment, participants answered demographic questions (gender, age, and education level), had the option to share remarks and were invited to take part in the lottery. Also, they were given the opportunity to leave their e-mail address if they were interested in receiving more information about the research and the results. The experiment took about 10 minutes.
Research materials of independent variable
For the manipulation of the purchase decision phase, participants chose a Bluetooth speaker which they could win by participating in the study. Bluetooth speakers were chosen for the same product characteristic reasons as the alarm clock radios in experiment 1. The product
25
was changed because now participants had to choose a product for themselves instead of for a good friend, in an attempt to make them more confident about their decision. Since music streaming is very popular and still gaining in popularity among consumers nowadays (GfK, 2016), it was expected that participants would be personally interested in this product.
The products, product descriptions, and specifications were taken from bol.com. They were modified following the same procedure as in experiment 1. The product information is included in Appendix E. The review titles and reviews were based on existing reviews at google.nl/shopping and bol.com. They were created and/or modified following the same procedure as in experiment 1. However, to make the review titles appear more veracious, valence star ratings and author names were added to the review titles. Also, in two
experimental conditions, review titles were accompanied with review useful votes. Negative review titles always had more review useful votes than positive review titles. The amount of useful votes for the positive review titles of both products ranged from 2 to 40 and the amount of useful votes for the negative review titles of both products ranged from 40 to 94. The amount of useful votes for the first product’s review titles were randomly chosen within these ranges. The amount of votes for the second product’s review titles were the same useful votes minus 1. For example, when a negative review of the first product received 60 useful votes, a negative review of the second product received 59 useful votes. To give an example of what it looked like, in Appendix F review titles for product one with useful votes are included. The complete reviews are included in Appendix G.
Measures
The same measures were used as in experiment 1, including decision confidence (M = 4.88, SD = 1.10, Cronbach's α = .80), selection of positive (M = 1.76, SD = 1.09) and of negative reviews (M = 3.24, SD = 1.09), relevance of positive (M = 5.29, SD = 1.05, Cronbach's α =
26
.90) and negative reviews (M = 5.48, SD = 1.03, Cronbach's α = .89), persuasiveness of positive (M = 5.03, SD = 1.08, Cronbach's α = .82) and negative reviews (M = 5.15, SD = 1.06, Cronbach's α = .79), decision involvement (M = 5.16, SD = 1.23, Cronbach's α = .85), yes/no question online review importance (yes = 91.5 %, n = 162), and scale item review importance (M = 5.85, SD = 1.02).
Results
Randomization checks
For a randomization check of gender across the experimental conditions, a crosstab with Chi-square test was performed. Gender was used as dependent variable and the conditions were used as independent variable. The analysis showed that the participant’s gender did not significantly differ between the purchase decision phases (χ² (3) = 1.92, p = .590).
For a randomization check of age, education level and the control variables of decision involvement and online review importance across the two purchase decision phases, One-way ANOVAs were conducted. For the yes/no control variable measuring online review
importance, a crosstab with Fisher’s Exact test was performed since 50% of the cells had an expected count below five. The participant’s individual characteristics of age, education level, decision involvement and online review importance were used as dependent variables and the experimental conditions were used as independent variable. The analyses showed that neither age (F (3,160) = 1.83, p = .145), nor education level (F (3,161) = .39, p = .764), nor decision involvement (F (3,161) = .37, p = .777), nor online review importance (F (3,161) = .48, p = .695), or yes/no online review importance (Fishers Exact p = .801) were significantly different between the different purchase experimental conditions. Therefore, none of these variables were included as covariates in the further analyses.
27 Hypotheses testing
The first two hypotheses (H1 and H2) and the last two hypotheses (H5 and H6) to be tested concerned the selection of review titles in and between different purchase decision phases and with or without review useful votes. A repeated measures ANOVA was used to test these hypotheses. The purchase decision phase and the review useful votes were used as the
between-subjects variables and the selection of positive and negative review titles was used as within-subjects variable. The ANOVA yielded a significant main effect of selected review titles (F (1, 146) = 62.58, p < .001, η² = 0.30). On average, participants selected more
negative review titles (M = 3.24, SD = 1.09) than positive review titles (M = 1.76, SD = 1.09). No significant interaction effect between purchase decision phase and review title selection was found (F (1, 146) = .22, p = .637). The purchase decision phase did not have an effect on the selection of reviews. Therefore H1, predicting that consumers in the post-purchase decision phase select more positive than negative review titles while consumers in the pre-purchase decision phase select an equal amount of positive and negative reviews is not confirmed. Also H2, predicting that consumers in the post-purchase decision phase select more positive and less negative review titles than consumers in the pre-purchase decision phase is not confirmed.
Also, no significant interaction effect was found between review useful votes and the selection of review titles (F (1, 146) = .00, p = .963). Review useful votes presence versus review useful votes absence did not lead to a difference in positive and negative review title selection. Therefore, H5,predicting that consumers in any purchase decision phase will select more reviews with a higher amount of review useful votes than reviews with a lower amount of review votes cannot be confirmed. To test if review useful votes moderate the effect of purchase decision phase on the selection of reviews, the three way interaction was interpreted. No interaction effect between purchase decision phase and review useful votes on review title
28
selection was found (F (1, 146) = 1.62, p = .205). Therefore, H6, predicting that the effect of the purchase decision phase on the selection of reviews is moderated by review useful votes is also not confirmed.
H3 and H4 concerned predictions regarding the perceptions of positive and negative reviews in and between the two different purchase decision phase conditions. Two repeated measures ANOVAs were conducted to test these hypotheses. The perceived review relevance and persuasiveness were used as within-subject variables and the purchase decision phase was used as the between-subject variable. The repeated measures ANOVAs did not yield a
significant main effect of review relevance (F (1, 163) = 2.60, p = .109) and review
persuasiveness (F (1, 163) = .76, p = .384). Overall, people perceived positive and negative reviews as equally relevant and persuasive. To examine differences within and between both purchase decision phases, interaction effects were interpreted. No significant interaction effects were found between the purchase decision phase conditions and the perceived review relevance (F (1,163) = .61, p = .436) or the perceived review persuasiveness (F (1,163) = .32, p = .576). People in both purchase decision phases perceived positive and negative reviews as equally relevant and persuasive. This partially confirms H3, predicting that consumers in the post-purchase decision phase perceive negative reviews as less relevant and less persuasive than positive reviews while consumers in the pre-purchase decision phase perceive negative and positive reviews as equally relevant and persuasive. H4, stating that consumers in the post-purchase decision phase perceive negative reviews as less relevant and less persuasive than consumers in the pre-purchase decision is not confirmed.
Exploration within the post-purchase decision condition group
The selection and perception of online reviews was again explored within the post-purchase decision condition itself as a function of decision confidence. First, a linear regression was
29
performed with the selection of positive and negative review titles as dependent variable and the decision confidence scale as the independent variable. No significant regression equation was found (F (1, 65) = 1.30, p = .258). Thus, someone’s decision confidence did not influence positive and negative review title selection in the second experiment.
Two other linear regression analyses were performed with the perceived relevance of positive and negative reviews as dependent variables and decision confidence as independent variable. The linear regression of perceived relevance of positive reviews yielded a significant result (F (1, 77) = 7.28, p = .009) with R² = .09. A positive significant effect was found
between someone’s decision confidence and the perceived relevance of positive reviews (β = .29, t = 2.70, p = .009). An increase of one point in decision confidence indicated an increased perceived relevance of positive reviews of .28 point. Someone’s decision confidence was not significantly related to the perceived relevance of negative reviews (F (1, 79) = .66, p = .132).
Two other linear regression analyses were performed to assess the impact of decision confidence on perceived review persuasiveness of positive and negative reviews. For
perceived relevance of positive reviews, a significant regression equation was found (F (1, 77) = 6.41, p = .013) with R² = .08. A positive effect was found between someone’s decision confidence and the perceived persuasiveness of positive reviews (β = .28, t = 2.53, p = .013). An increase of one point in decision confidence indicated an increased perceived
persuasiveness of positive reviews of .29 point. The perceived persuasiveness of negative reviews was marginally significant related to someone’s decision confidence (F (1,79) = 1.09, p = .057) with R²= .07. A positive trend was found between someone’s decision confidence and the perceived persuasiveness of negative reviews (β = .21, t = 1.94, p = .057). An increase of one point in decision confidence indicated an increased perceived persuasiveness of
30 Conclusion experiment 2
In sum, no support was found for H1, H2, H3, H5, and H6. With regard to H1, no support was found that consumers in the post-purchase decision condition selected more positive than negative review titles while consumers in the pre-purchase decision phase selected an equal amount of positive and negative reviews. With regard to H2, no support was found that consumers in the post-purchase decision phase selected more positive and less negative review titles than consumers in the pre-purchase decision phase. Instead, all consumers selected more negative than positive review titles, just as the result in experiment 1. With regard to H3, no support was found that people in the post-purchase decision phase perceived negative reviews as less relevant and less persuasive than people in the pre-purchase decision phase. Instead, all participants perceived negative and positive reviews as equally relevant and persuasive. This partially confirms H4, predicting that consumers in the post-purchase
decision phase perceive negative reviews as less relevant and less persuasive than positive reviews while consumers in the pre-purchase decision phase perceive negative and positive reviews as equally relevant and persuasive. As for H5 and H6, it could not be confirmed that review useful votes have an impact on the selection of positive or negative review titles.
When exploring differences between participants within the post-purchase decision condition itself, some expected outcomes were obtained: although decision confident
consumers did not select more positive and less negative reviews compared to consumers who were uncertain about their decision, the more confident consumers were, the more relevant and the more persuasive they perceived positive reviews.
General discussion
This paper focussed on the effect of purchase decision phases and review useful votes on the selection and perception of reviews. This study contributes to the literature by increasing the
31
existing knowledge on individual characteristics that influence consumers’ perception and processing of eWOM information, which has been identified as a relevant gap in eWOM research (King, Racherla, & Bush, 2014). In the experiments conducted in this paper, not only opinions, but also actual behaviour was measured by asking people to select five out of twelve review titles they thought were interesting. Also, it combines research on consumer
characteristics with research on online review cues by examining the disrupting effect of review useful votes on selective exposure. The results of this study therefore help researchers as well as practitioners to gain a better understanding of consumers’ online review processing.
The two experiments in this paper demonstrated that overall, consumers selected more negative than positive review titles to read, regardless the purchase decision phase they were in. Although this was not predicted, previous research on review engagement has found comparable findings in terms of attention to and engagement with negative versus positive information (Carlson, Guha, & Daniels, 2011; Sen, & Lerman, 2007). They explained this bias in terms of salience of negative information (Sen & Lerman, 2007). Usually, one’s social environment contains more positive than negative cues. This makes negative cues counter normative (Kanouse, & Hanson, 1972). As a consequence, the negative cues that appear tend to be more salient and attract more attention than positive cues (Ahluwalia, 2000; Kanouse, & Hanson, 1972). Only results from experiment 1 suggest that consumers can have a
confirmation bias in the selection of review titles, caused by a high decision confidence. However, this effect did not occur in experiment 2. Possibly age played a role in the selection of review titles, since participants in experiment 1 were on average about twenty years older than participants in experiment 2. On average, participants in experiment 1 thus belonged to generation X while participants in experiment 2 overall belonged to generation Y. Generation Y members are known to ascribe more value to others’ opinions and feedback in online media than older generations (Bolton, et. al., 2013). Therefore, it is possible that even when they are
32
confident about their decision, they will always be more interested in negative evaluations provided by others than older generations. It would be interesting to examine this further in future research.
The results of experiment 1 indicate that differences exist in the perceptions of positive and negative reviews between and/or within different purchase decision phases in one of the experiments. In experiment 1, support was found that consumers in the post-purchase decision perceive negative reviews as less relevant and less persuasive than people in the pre-purchase decision phase. This was also predicted based on the literature. Another result of the first experiment, although not predicted based on the literature, was that people in the pre-purchase decision phase perceived negative reviews as more relevant and persuasive than positive reviews. On the contrary, people in the post-purchase decision phase perceived negative and positive reviews as equally relevant and persuasive. Apparently, prior preferences of
consumers in the post-purchase decision phase enabled them to effectively discount
information in the negative reviews (Ahluwalia, 2000; Sen & Lerman, 2007). This resulted in an equal relevance and persuasiveness perception of positive and negative reviews. Therefore the first experiment indicated that instead of having a confirmation bias in the post-purchase decision phase, people rather have a negativity bias in the pre-purchase decision phase. This bias is rectified in the post-purchase decision phase. Notable is that these differences did again not occur in experiment 2 for generation Y participants. No significant differences in
perceived relevance or persuasiveness in or between different purchase decision phases were found in experiment 2. Only decision confident consumers in both experiments showed a bias in their perception of positive reviews. These were perceived as more relevant and more persuasive when consumers were more confident.
Although the review useful votes did not have a significant effect in this study, no real conclusions can be drawn regarding this review cue. Since participants in experiment 2
33
selected in general double as much negative reviews compared to positive reviews, regardless of the review useful votes, it is impossible to conclude that useful votes are not effective. Especially because recent research on online review cues found through interviews and a survey that consumers themselves indicate that they use review useful votes as information cues to select reviews more easily. Therefore, further research is needed to test whether review useful votes have an effect on review selection or not.
Limitations and implications
A limitation of the current study is that the results cannot be generalized to all product
categories. Only the selection and perception of reviews for high-involvement audio products were studied. Therefore, the results of this study only apply to this product category. It is possible that the selection and perception of reviews is different for other product types. For example, people might have a stronger confirmation bias for experience products compared to search products, because opinions regarding experience products or services are more
subjective. For example, whereas one person might detest a movie, it might be someone else’s favourite. Future research may expand the results of the current by examining the selection and perception of online reviews for other product categories as well.
Another limitation concerns the sample size of the second experiment. When starting the analysis, it appeared that quite some participants skipped the review title selection question. Because of time restrictions, it was impossible to collect more data to reach the desired amount of at least 40 participants per condition. This limited the amount of useful data and caused that two experimental conditions only had 30-35 participants for this item. This could have influenced the results of this experiments.
The results of this study have implications for researchers as well as webshop
34
as one of the most powerful persuasive tools of the 21st century (King, Racherla, & Bush, 2014; Mafael, Gottschalk & Kreis, 2016). However, the fact that the consumer’s cognitive stage - and especially ones decision confidence - affects how relevant and persuasive he or she perceives reviews should be taken into account by consumer researchers and webshop marketers studying the effects of product reviews. Furthermore, they should take into account that age differences might cause different selection and perception patterns. Overall, the results of this study suggest that managers should put in effort to make consumers want their product before they are exposed to products reviews. Being in the post-purchase decision phase makes people perceive positive and negative reviews more equal in terms of relevance and persuasiveness. Furthermore, once people are confident about their decision, they also assign more value to positive reviews. This increases the chance consumers will eventually buy a product. Therefore, managers and consumer behaviour researchers may also be interested in examining the best way to present products (e.g. product descriptions, product features, and the use of images) in an online environment in future research.
References
Ahluwalia, R. (2000). Examination of psychological processes underlying resistance to persuasion. Journal of Consumer Research, 27(2), 217-232.
Bimber, B., & Davis, R. (2003). Campaigning online: The Internet in U.S. elections. New York, NY: Oxford University Press.
Bojanic, D. C., & Warnick, R. B. (2012). The role of purchase decision involvement in a special event. Journal of Travel Research, 51(3), 357-366.
35
Gruber, T., Komarova Loureiro, Y., & Solnet, D. (2013). Understanding Generation Y and their use of social media: a review and research agenda. Journal of Service
Management, 24, 245-267.
BrightLocal. (2016). Local Consumer Review Survey 2016. Retrieved from https://www.brightlocal.com/learn/local-consumer-review-survey/
Brock, T. C., & Balloun, J. L. (1967). Behavioral receptivity to dissonant information.
Journal of Personality and Social Psychology, 6(4), 413-428.
Carlson, K., Guha, A., & Daniels, R. M. (2011). The ratings paradox: Why we prefer reading negative reviews, but then subsequently rate these reviews as less useful. In: D. W. Dahl et. al. (Eds.). NA-Advances in Consumer Research. (Vol. 38). Duluth, MN: Association for Consumer Research.
Cialdini, R.B., Reno, R.R., & Kallgren, C.A. (1990). A focus theory of normative conduct. Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social Psychology, 58, 1015-1026.
Dtg. (2015, 9 November). Klantbeoordelingen van grote invloed op online shoppen. Retrieved from
https://www.dtg.nl/over-ons/nieuws-en-pers/persberichten/onderzoek-klantbeoordelingen-grote-invloed-online-shoppen.
Feather, N. T. (1963). Cognitive dissonance, sensitivity, and evaluation. Journal of Abnormal and Social Psychology, 66, 157-163.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.
Fischer, P., Fischer, J., Weisweiler, S., & Frey, D. (2010). Selective exposure to information: How different modes of decision making affect subsequent confirmatory information processing. British Journal of Social Psychology, 49(4), 871-881.
36
integrative model. Current Directions in Psychological Science, 19, 384-389. Fischer, P., Greitemeyer, T., & Frey, D. (2008). Self-regulation and selective exposure: The
impact of depleted self-regulation and resources on confirmatory information processing. Journal of Personality and Social Psychology, 94, 382-395.
Fischer, P., Schulz-Hardt, S., & Frey, D. (2008). Selective exposure and information quantity: How different information quantities moderate decision makers’ preference for
consistent and inconsistent information. Journal of Personality and Social Psychology, 94(2), 231-244.
Fransen, M. L., Smit, E. G., & Verlegh, P. W. J. (2015). Strategies and motives for resistance to persuasion: An integrative framework. Frontiers in Psychology, 6, 1201–1221. Frey, D. (1986). Recent research on selective exposure to information. In L. Berkowitz (Ed.).
Advances in Experimental Social Psychology (Vol. 19, pp. 41–80). New York: Academic Press.
GfK. (2016, 31 August). Bluetooth and streaming solutions set the tone in the audio market. Retrieved from http://www.gfk.com/insights/press-release/ifa-2016-audio-en/. Gottschalk, S. A., & Mafael, A. (2017). Cutting through the online review jungle –
investigating selective eWOM processing. Journal of Interactive Marketing, 37, 89-104.
Hart, W., Albarracin, D., Eagly, A. H., Brechan, L., Lindberg, M. J., & Merril, L. (2009). Feeling Validated versus being correct: A meta-analysis of selective exposure to information. Psychological Bulletin, 155, 555-588.
Jacobson, R.P., Mortensen, C.R., & Cialdini, R.B. (2011). Bodies obliged and unbound: differentiated response tendencies for injunctive and descriptive social norms. Journal of Personality and Social Psychology, 100, 433-448.
37
information search after preliminary decisions: An expansion of dissonance theoretical research on selective exposure to information. Journal of Personality and Social Psychology, 80 , 557-571.
Kanouse, D. E., & Hanson, L. R. (1972). Negativity in evaluations. In: E. E. Jones et. al. (Eds.), Attribution: Perceiving the causes of behaviour. Morristown, NJ: General Learning Press.
King, R. A., Racherla, P., & Bush, V. D. (2014). What we know and don’t know about online Word-Of-Mouth: A review and synthesis of the literature. Journal of Interactive Marketing, 28(3), 167-183.
Liang, Y. J. (2016). Reading to make a decision or to reduce cognitive dissonance? The effect of selecting and reading online reviews from a post-decision context. Computers in Human Behavior, 64, 4630471.
Mafael, A., Gottschalk, S. A., & Kreis, H. (2016). Examining biased assimilation of brand- related online reviews. Journal of Interactive Marketing, 36, 91-106.
Martin, C. L. (1998). Relationship marketing: A high-involvement product attribute approach. Journal of Product & Brand Management, 7(1), 6-26.
Mudambi, S. M., & Schuff, D. (2010). What makes a helpful online review? A study of customer reviews on Amazon.com. MIS Quarterly, 34(1), 185-200.
Munro, G. D., Ditto, P. H., & Lockhart, L. K. (2002). Biased assimilation of sociopolitical arguments: Evaluating the 1996 U.S. Presidential debate. Basic and Applied Social Psychology, 24(1), 15-26.
Park, D., & Lee, J. (2008). eWOM overload and its effect on consumer behavioral intention depending on consumer involvement. Electronic Commerce Research and
Applications, 7(4), 386-398.
