What to Choose? A Study on User’s Choice for an Online Review System.

(1)

1

What to Choose? A Study on

User’s Choice for an Online

Review System.

Student: Lotte de Vogel

Student nr.: s4236297 Date: 19-06-2017

Supervisor: Bas Hillebrand

(2)

2

Abstract

This study examined which factors determine user’s choice for a professional online review system or a consumer online review system. The study focussed on two opposites, namely if consumers choose an online review system because they want to read reviews of people who they consider to be similar to themselves or because they want to distinguish themselves from others. The research question was: ‘To what extent do user’s expertise and

status motives have an influence on user’s choice for an online review system (consumer online review system vs. professional online review system) in the context of restaurants?.

In order to answer this research question, an experiment was conducted in which status motives were manipulated. In line with what was expected based on the similarity-attraction theory (Byrne, 1961), for women, user’s expertise had an effect on user’s choice for an online review system. Women with higher expertise were more inclined to choose a professional review system, while women with lower expertise were more inclined to choose a consumer review system. For men, however, no significant result was found for user’s expertise on user’s choice for an online review system.

In contrast to what was expected based on the upward mobility theory (Bourdieu, 1984), for women, status motives did not influence user’s choice for an online review system. For men, however, status motives did have a significant influence on user’s choice for an online review system. Men primed with status-motives were more inclined to choose a professional review system, while men who were not primed with status-motives were more inclined to choose a consumer review system.

From a theoretical perspective, this study contributed to the literature on online reviews, since this is the first study that focussed on user’s choice for an online review system instead of on individual reviews. User’s choice for an online review system is an important variable, because the choice for an online review system determines which online reviews consumers see and eventually may influence purchase decisions of consumers. In addition, it is more realistic than past research since consumers first choose a system before they choose a restaurant. From a managerial perspective, this study provides restaurant owners and owners of review systems with information about what type of consumers are likely to use which type of online review system.

(3)

3

1. Introduction

1.1 Introduction

Since the emergence of the internet, the world is increasingly becoming a global village. In this global village people can easily have access to online reviews written by people from all over the world via online review systems. Especially for experience goods, which are products or services of which the product quality is difficult to assess in advance (Nelson, 1970), online reviews are important (Cox & Kaimann, 2015). For these experience goods, online reviews can serve as an indicator of quality and in that way reduce uncertainty.

Nowadays, online review systems for all types of product and services exist. Some of these online review systems are based on professional reviews, whereas other online review systems are based on consumer reviews. Consumer review systems consist of consumer reviews which are written by individual consumers, without any formal authority, who share their experiences with a product or service with other consumers (Flanagin & Metzger, 2013). Professional review systems consist of reviews written by experts, with formal authority, who often acquired their expertise via education or training.

For restaurants, both types of review system exist in the Netherlands. A popular online review system based on professional reviews is Lekker. Each year Lekker composes a list of the top 500 restaurants in the Netherlands (Lekker, 2017). This top 500 is based on the evaluations of 68 professional reviewers with a background in hospitality. Each restaurant in the top 500 is rated on 60 evaluation criteria, such as quality of the dishes and wine advice given. Another popular online review system is IENS, which is a review system based on consumer reviews. Each month more than 2 million unique visitors visit the IENS website to search for a restaurant (Iens, 2017). In addition, 200.000 new reviews are placed by consumers on IENS on a yearly basis. Since online reviews are especially important for experience goods such as restaurants and for restaurants both professional review systems and a consumer review systems exist in the Netherlands, this study was conducted in a restaurant context.

1.2 Problem statement

Past research has mainly focussed on individual reviews. While several studies have investigated individual consumer reviews (Awad & Ragowski, 2008; Bailey, 2005; Chevalier & Mayzlin, 2006; Clemons et al., 2006; Dabholkar, 2006; Goldsmith & Horowitz, 2006; Hennig-Thurau & Walsh, 2003; Kim et al., 2011; Lim et al., 2006; Sweeney et al., 2006; Ye et

(6)

6

al., 2009; Zhu & Zhang, 2010), other studies have examined individual professional reviews (Basuroy, Ravid & Hall, 2014; Legoux, Larocque, Laporte, Belmati & Boquet, 2016; Reinstein & Snyder, 2005). In addition, some studies have compared individual consumer reviews with individual professional reviews (Cox & Kaimann, 2015; Chiou et al., 2013; Huang & Chen, 2006; Plotkina & Munzel, 2016; Smith et al., 2005; Zhang et al., 2010; Zhou & Duan, 2010). Some of these studies on individual reviews have focussed on the consequences for users of online reviews and found a positive effect on sales (Chevalier & Mayzlin, 2006; Clemons, Gao & Hitt, 2006; Cox & Kaimann, 2015; Reinstein & Snyder, 2005; Ye, Law & Gu, 2009; Zhu & Zhang, 2010; Zhou & Duan, 2010), purchase intention (Chiou, Hsiao, & Fang, 2013; Huang & Chen, 2006; Plotkina & Munzel, 2016), online popularity (Zhang, Ye, Law & Li, 2010) and trust (Awad & Ragowski, 2008; Lim, Sia, Lee, & Benbasat, 2006).

In addition, other studies that focussed on individual reviews have examined the antecedents of online reviews and found that users mainly read reviews to reduce search efforts and perceived risk (Dabholkar, 2006; Goldsmith & Horowitz, 2006; Hennig-Thurau & Walsh, 2003; Kim, Baloglu & Mattila, 2011; Sweeney, Soutar & Mazzarol, 2006), for social assurance (Bailey, 2005; Hennig-Thurau & Walsh, 2003; Kim et al., 2011) and to maximize consumption benefits (Goldsmith & Horowitz, 2006; Kim et al. 2011; Sweeney et al., 2006).

Although the antecedents and consequences for users of individual reviews have been examined by several studies, to my knowledge there is no study that has examined user’s choice for an online review system. User’s choice for an online review system is, however, an important variable that needs to be studied since the choice for an online review system influences which individual reviews consumers see and eventually can influence the choice for a restaurant.

1.3 Research question

The choice for an online review system might depend on personal characteristics of the user. This study examines two opposites. On the one hand, some users might choose an online review system, because they want to read reviews of other consumers who they consider to be similar to themselves with regard to expertise. On the other hand, some users might choose an online review system, because they want to acquire a higher social status and in that way distinguish themselves from other consumers.

Although user’s expertise has been studied in the context of online reviews (Kim et al., 2011; Park & Kim, 2008), the effect of user’s expertise on user’s choice for an online review

(7)

7

system has not been investigated yet. User’s expertise might, however, be an important antecedent of user’s choice for an online review system. Past research found that user’s expertise has an effect on user’s motivation to read online reviews (Kim et al., 2011) and on product choice (Moorman, Diehl, Brinberg & Kidwell, 2014). Since user’s expertise has an effect on user’s motivation to read online reviews and on product choice, user’s expertise might also have an effect on user’s choice for an online review system.

Additionally, the choice of a review system may depend on status motives. Past research found that respondents with status motives chose a different product than respondents who did not have status motives (Griskevicius, Tybur & van den Bergh, 2010). Since status motives influence product choice, status motives might also influence user’s choice for an online review system. In order to gain knowledge about the effect of user’s expertise and status motives on user’s choice for an online review system, the following research question was formulated:

‘To what extent do user’s expertise and status motives have an influence on user’s choice for an online review system (consumer online review system vs. professional online review system) in the context of restaurants?’

1.4 Theoretical contribution

Firstly, this research adds on the literature on online reviews since this is the first study that focusses on user’s choice for online review system instead of the antecedents and consequences of individual online reviews which have been studied extensively. User’s choice for an online review system is an important variable, because the choice for an online review system determines which online reviews consumers see and eventually may influence purchase decisions of consumers. In addition, it is more realistic than past research that mainly focussed on individual reviews, since consumers first choose a system before they choose a restaurant. Secondly, this research adds to the literature on expertise and status. To my knowledge, this is the first study that examines whether user’s expertise and status motives have an effect on user’s choice for an online review system. User’s expertise and status motives are important variables to examine, because both variables were found to influence product choices (Griskevicius et al., 2010; Moorman et al., 2014) and therefore these variables might also influence the choice for an online review system which in turn might influence the choice for a restaurant.

(8)

8

1.5 Mangerial implications

This research can provide insight in which factors influence the choice for an online review system. This knowledge can help restaurant owners to determine the strategy of the restaurant with regard to online reviews. If a restaurant owner, for example, knows that the target group of the restaurant is more inclined to choose a consumer review system, then it is important to motivate consumers who have visited the restaurant to post a positive review on an online consumer review system. Since restaurant owners only have limited resources, such as time and money, insights in user’s choice for an online review system can help to determine the strategy of the restaurant with regard to online reviews. If the online review strategy of the restaurant fits the preference for a certain type of online review system of the target group of the restaurant, the restaurant owner can serve the target group better and this will probably increase satisfaction and eventually increase purchase intentions of the target group. Higher purchase intentions can eventually increase the actual purchase behaviour (Ajzen, 1985) and in that way improve the financial performance of the restaurant.

In addition, the results of this study are relevant for owners of review systems since this study gives insight in what factors influence user’s choice for an online review system. If, for example, an owner of a review system knows that consumers with high expertise are more likely to visit their system, than they can serve this particular target group better.

1.6 Structure of the report

In chapter 2 an overview of the relevant literature on online reviews will be given and the reasoning for hypothesis will be explained. In chapter 3, the methodology used in the study will be discussed. Next, in chapter 4, the results will be discussed. Finally, a conclusion and discussion will be given in chapter 5.

(9)

9

2. Theoretical Background

Word-of-mouth can have a large influence on consumer purchase decisions (Chevalier & Mayzlin, 2006; Clemons et al., 2006; Ye et al., 2009; Reinstein & Snyder, 2005; Zhu & Zhang, 2010). Before the internet existed, the person who spread the word-of-mouth only influenced people that were close to him, such as family and friends. However, since the rise of the internet consumers can easily have access to online reviews, written by professionals and consumers from all over the world, via online review systems.

2.1 Previous studies on the users of consumer and professional online

reviews

As can be seen from table 1, past research on users of online reviews has mainly focused on individual reviews. Whereas some studies have investigated individual consumer reviews (Awad & Ragowski, 2008; Bailey, 2005; Chevalier & Mayzlin, 2006; Clemons et al., 2006; Dabholkar, 2006; Goldsmith & Horowitz, 2006; Hennig-Thurau & Walsh, 2003; Kim et al., 2011; Lim et al., 2006; Sweeney et al., 2006; Ye et al., 2009; Zhu & Zhang, 2010), other studies have examined individual professional reviews (Basuroy, Ravid & Hall, 2014; Legoux, Larocque, Laporte, Belmati & Boquet, 2016; Reinstein & Snyder, 2005). There are also studies that have compared individual consumer reviews with individual professional reviews (Cox & Kaimann, 2015; Chiou et al., 2013; Huang & Chen, 2006; Plotkina & Munzel, 2016; Smith et al., 2005; Zhang et al., 2010; Zhou & Duan, 2010).

In addition, past research has focussed on the consequences of individual reviews and found that online reviews have, for example, an effect on sales (Basuroy et al., 2014; Chevalier & Mayzlin, 2006; Clemons et al., 2006; Cox & Kaimann, 2015; Reinstein & Snyder, 2005; Ye et al., 2009; Zhou & Duan, 2010; Zhu & Zhang, 2010), purchase intention (Chiou et al. 2013, Huang & Chen, 2006; Plotkina & Munzel, 2016) and online popularity (Zhang et al., 2010).

Next to the consequences of individual reviews, other studies have examined the antecedents for users of individual online reviews and examined why users read online reviews. These studies found that users mainly read online reviews to reduce search efforts and perceived risks (Dabholkar, 2006; Goldsmith & Horowitz, 2006; Hennig-Thurau & Walsh, 2003; Kim et al., 2011; Sweeney et al., 2006), social assurance (Bailey, 2005; Henning-Thurau& Walsh, 2003; Kim et al., 2011) and maximize consumption benefits (Goldsmith & Horowitz, 2006; Kim et al., 2011; Sweeney et al., 2006).

(10)

10

reviews, this present study investigates why users choose a particular online review system. This study examines two opposites, namely if users choose an online review system because they want to read reviews of other consumers who they consider to be similar to themselves or because they want to distinguish themselves from others. Some users might choose a particular online review system, because they believe that the writers of the reviews that are posted on the system are similar to themselves with regard to expertise. In contrast, other users might choose an online review system because they want to acquire a higher social status and as a result distinguish themselves from others. The conceptual model can be found in figure 1.

Figure 1. Conceptual model

User's choice for an online review system

(professional review online system vs. consumer online review system) User's exerptise Status motives

(11)

11

Table 1. Previous studies on the users of consumer and professional online reviews

Individual review level Review system level

Consequences for users of online reviews (What is the effect of online reviews?)

Antecedents for users of online reviews (Why do users read online reviews?)

Antecedents of user’s choice for an online review system (Why do users choose for a review system?) Consumer reviews • Sales: Chevalier & Mayzlin, 2006; Clemons et al.,

2006; Ye et al., 2009; Zhu & Zhang, 2010 • Trust: Awad & Ragowski, 2008; Lim et al., 2006

• Reduce search efforts and perceived risks: Dabholkar, 2006; Goldsmith & Horowitz, 2006; Hennig-Thurau & Walsh, 2003; Kim et al., 2011; Sweeney et al., 2006

• Social assurance: Bailey, 2005; Hennig-Thurau & Walsh, 2003;

Kim et al., 2011

• Maximize consumption benefits: Goldsmith & Horowitz, 2006; Kim et al. 2011; Sweeney et al., 2006

Professional reviews • Sales: Basuroy et al., 2014; Reinstein & Snyder,

2005

• Exhibitor’s decision to withdraw a movie or not: Legoux et al., 2016

• Marketing activities: Basuroy et al., 2014

Consumer vs. professional reviews

• Sales: Cox & Kaimann, 2015; Zhou & Duan, 2010 • Purchase intention: Chiou et al. (2013): Huang &

Chen, 2006; Plotkina & Munzel, 2016 • Online popularity: Zhang et al., 2010

(12)

12

2.2 Choice for online review system

Choice can be defined as ‘making a statement about the individual’s taste and values’ (Mudambi & Schuff, 2010, p. 190). Choice for an online review system in this study is defined as the dichotomous choice between an online consumer review system or an online professional review system. A consumer review system in this study is defined as a review system that consists of consumer reviews which are written by individual consumers, without any formal authority, who share their experiences with a product or service with other consumers (Flanagin & Metzger, 2013). A professional review system in this study is defined as a review system that consists of reviews written by experts, with formal authority, who often acquired their expertise via education or training.

2.3 User’s expertise

Users can differ in their level of expertise and knowledge about a product or service (Chen & Xie, 2008). Some users have high knowledge about a product or service and are therefore considered to be experts, while other users only have limited knowledge about a product or service and are therefore seen as novices. In this study, user expertise is defined as ‘a person’s

embodied skill or expertise that is acquired through past experiences, formal or informal training, or education’ (Luo & Toubia, 2015, p. 102).

Several studies have examined the role of user’s expertise in the context of online reviews (Kim et al., 2011; Park & Kim, 2008). User’s expertise was found to have an effect on purchase intention (Park & Kim, 2008). This relationship was, however, moderated by type of review. Experts had higher purchase intention when they were exposed to reviews that focussed on attributes, while novice had higher purchase intention when they read reviews that focussed on benefits. Furthermore, user’s expertise was found to influence user’s motivation to read online reviews (Kim et al., 2011). Users with medium levels of expertise were found to have the highest motivation to read online reviews for social reassurance.

According to the similarity-attraction theory people are attracted to others who they consider to be similar to themselves (Byrne, 1961). The similarity-attraction theory can also be extended to electronic word-of-mouth (Rosario, Sotgiu, de Valck & Bijmolt, 2016). For social media platforms the effect of electronic word-of-mouth is higher when users consider themselves similar to the senders of the electronic word-of-mouth.

Based on the similarity attraction theory, it can be expected that users with a high level of expertise are more attracted to online reviews of professionals with a similar level of expertise. Users probably believe that professionals with similar levels of expertise have the

(13)

13

same high standards and find the same criteria important when choosing a restaurant. As a result, the users have more trust in reviews written by professionals and therefore the users are probably more inclined to choose a professional review system. Users with high expertise know that they have higher expertise than average users who post reviews on consumer review systems. Users with high expertise might believe that they also have higher standards and other evaluation criteria than average users who post reviews on consumer review systems. As a result, users with high expertise have less trust in the reviews written by consumers and therefore they might also be less inclined to choose a consumer review system.

On the other hand, users with a low level of expertise are more likely to be attracted to reviews written by other consumers since they consider other consumers as similar to themselves with regard to expertise. Due to this similarity, users might believe that the consumers who wrote the reviews finds similar things important when it comes to restaurants and therefore they trust consumer reviews more. As a result, it can be hypothesised that users with a low level of expertise are more inclined to choose consumer review systems. Users with low expertise are probably less inclined to choose professional review systems, since they believe that these professionals have more expertise and higher standards when it comes to choosing a restaurant. Consequently, the following hypothesis for the effect of user’s expertise on user’s choice for an online review system is formulated:

H1: If the expertise of the user is high (low), then the user is more (less) inclined to choose a professional review system

2.4 Status motives

Consumers who aim to acquire a higher status want to distinguish themselves from others and consequently may choose another type of online review system, than consumers for who status is not important. In this study, status motives are defined as ‘the desire to gain prestige, respect,

or power over others’ (Griskevicius et al., 2009).

Several studies have examined the role of status motives (Griskevicius et al., 2009, Griskevicius et al., 2010; Yang & Matilla, 2013). Past research showed that status motives can easily be manipulated by using a status-related story (Griskevicius et al., 2009; Griskevicius et al., 2010). Status motives were found to have an effect on product choice (Griskevicius et al. 2010). Activating status motives can lead consumers to prefer green products over more luxurious non-green products when shopping in public. A reason for this preference is that shopping green products enhanced the status of the consumer since it helped consumers to be seen as prosocial instead of proself. Furthermore, status was found to have an effect on

(14)

14

consumer’s word-of-mouth intentions (Yang & Matilla, 2013). Consumers for who status was important were found to be more inclined to talk about their luxury goods purchases than consumers for who status was not important. Since consumers who are activated with status motives are found to have different purchase decisions and word-of-mouth behaviour than consumers who were not activated with status motives, it can also be expected that activating status motives influences user’s choice for an online review system.

According to the upward mobility theory of Bourdieu (1984), some people aspire to move up the social ladder in society. People with high status motives are more attracted to groups, roles, people and organizations that they consider as high status (Anderson, Hildreth, & Howland, 2015). Users who want to move up in society are looking for restaurants that fit with the high status that they aspire. Users with status motives may have an ideal self-image of who they would like to become in their mind. The ideal self-image can be defined as ‘the

imagination of ideals and goals related to what a person believes that he or she would like to be or aspire to become’ (Malar, Krohmer, Hoyer & Nyffenegger, p.36). Consumers tend to

choose products that fit with their ideal self-image. Users can achieve this fit by choosing products and services that fit with their ideal self-image. Users with status motives therefore might be more inclined to choose a professional review system since users consider the culinary journalist as having more status than ordinary consumers who write reviews on consumer review systems. As a result, professional review systems fit better with their ideal self-image than consumer review systems.

It can be expected that users for who status motives are activated, are more inclined to choose a professional review system than users for who status motives are not activated. Consequently, the following hypothesis was formulated:

H2: If a user is (not) primed with status motives, then the user is more (less) inclined to choose for a professional review system

(15)

15

3. Methodology

3.1 Research strategy and design

In order to examine the effect of user’s expertise and status motives on user’s choice for an online review system, a one-factor experiment was conducted with a between-subjects design. An experiment was chosen as research strategy since an experiment is an appropriate method to manipulate variables and investigate what the effect of this manipulation on an outcome variable is (Field, 2009). The factor that was manipulated is status motives. This factor had two levels, namely status motives and no status motives. Since in a between-subjects design respondents are only exposed to one experimental condition (Field, 2009), respondents were only exposed to one level. Respondents in the experimental group were exposed to a status-related story that was found to manipulate status motives in past research (Griskevicius et al., 2009, Griskevicius et al., 2010, Knechel & Leiby, 2016). In addition, there were two control groups who were not primed with status motives. Respondents in the first control group were exposed to a non-status related story and respondents in the second control group were not exposed to any text. Two control groups were used to assure that potential results were not biased by some particular aspect of the control story. User’s expertise was measured instead of manipulated since measuring expertise with a scale is a good indication of the expertise of consumers and how confident consumers are about their expertise (Park & Lessig, 1981). In addition, it would have been difficult to manipulate knowledge about restaurants since knowledge structures develop over time with increased experience with the domain (Sujan, 1985). Therefore, it was decided to measure user’s expertise with regard to restaurants.

Before the real experiment was conducted, a pre-test was done with 12 respondents using the thinking aloud procedure. The thinking aloud procedure was used because this method is appropriate to get insides in how people solve problems and in that way gain in-depth knowledge about the thoughts of participants (Ericsson & Simon, 1984).

(16)

16

3.2 Outline experiment

As can be seen from figure 2, the experiment started with the status motives manipulation and a manipulation check was conducted to check whether the status-related text really manipulated status motives. After that respondents were introduced to two online review systems, namely a professional review system and a consumer review system. After they read the information about the two systems, they were asked to choose which system they would choose to search for a restaurant. Next, user’s expertise was measured and several control questions were asked. Finally, a control question was asked about the status-related or non-status related text to check whether the respondents read the text carefully.

Figure 2. Outline of the experiment

3.3 Procedure

An online questionnaire in Qualtrics was created. The questionnaire can be found in Appendix 1. Except for the status motives manipulation, the questionnaires were exactly the same. Respondents were randomly assigned over the three conditions. Randomization was used to minimize the unsystematic variation (Field, 2009). Only adults who were 18 years or older could participate in the experiment. This criteria was used because adults are more likely to make reservations at restaurants and search for restaurants online than children who rarely make reservations at restaurants and therefore are also less likely to search for a restaurant online.

The questionnaire was send to the respondents within the personal network of the researcher via a link in an e-mail and via Facebook and LinkedIn. In this way, respondents

(17)

17

could decide themselves when they had time to fill in the questionnaire on their computer or phone. To make sure that there was enough variation in user’s expertise, respondents that have a hospitality education and respondents that do not have a hospitality education were asked. Firstly, respondents in the personal network of the researcher of who the researcher was aware that they have a hospitality background were asked. Experts are likely to know other experts (Lilien, Morrison, Searls, Sonnack & von Hippel, 2002). Therefore, the snowball sampling technique was used to get access to other people with knowledge about restaurant who would be willing to fill in the questionnaire (Biernacki & Waldorf, 1981). Also respondents in the personal network of the researcher who rarely go to restaurants and have limited knowledge about restaurants and respondents with average knowledge about restaurants were asked. In this way, it could be expected that there was enough variation between respondents on user’s expertise about restaurants.

The mean duration for the experimental group who read the status-related text was 10.87 minutes and for control group who read the non-status related text was 11.05. The mean duration for the control group who did not read any text was 7.40 minutes.

3.4 Ethics

In the introduction text of the experiment, it was emphasised that the questionnaire can be filled in anonymously, the data will be treated confidentially and that there are no wrong answers. Respondents were free to withdraw from the experiment at any time. At the end of the experiment, respondents could fill in their e-mail address if they want to be informed about the results. Respondents participated in the experiment on a voluntary basis and were told that they could win a bol.com cheque of 30 euro when they completed the experiment.

3.5 Respondents

In total, 146 respondents took part in the experiment. From these 146 respondents, 3 respondents were not included for further analysis, because they did not answer the control question about the manipulation text correctly. Therefore, it cannot be assumed that they read the status-related or non-status-related text carefully. Furthermore, one subject was younger than 18 years old and was also not included for further analysis. Finally, 3 respondents who saw a status-related or non-status related text conducted the experiment in more than 30 minutes while the average time for the status-related text was 10.87 minutes and for the non-status related text 11.05 minutes. These respondents were not included because the relatively long

(18)

18

time that the respondent took to fill in the questionnaire, could have influenced the effect that the status-related or non-status related text have had on the respondents when choosing for an online review system. If a respondent, for example, did something else in between the experiment, this could have influenced the effect that the manipulation of status motives had on user’s choice for an online review system. After these respondents were removed from the data set, 139 respondents were left for further analysis. The experimental group who saw the status-related text consisted of 68 respondents. The control group who saw the non-status related text consisted of 33 respondents and the control group who did not see a text consisted of 38 respondents. Since all groups consisted of at least 30 respondents, a normal distribution could be assumed (Hair, Black, Babin & Anderson, 2014).

As can be seen from table 2, 71.9 percent of the sample was women and the mean age was 35 years old (sd = 15.72, range = 49). The most frequently mentioned educational level was WO. Most respondents are going 1 to 2 times per month to a restaurant. In addition, most respondents spend on average between 16 and 30 euro per person when they go to a restaurant. In general, respondents were more familiar with consumer review systems (M = 4.93, sd = 1.40, range = 6) than with professional review systems (M = 3.27, sd = 1.47, range = 5.50).

(19)

19

Table 2. Descriptive statistics sample

Variables Descriptives Gender Women 71.9 % Men 28.1 % Age Mean 35.24 SD 15.72 Min 18 Max 67 Education WO 44.6 % HBO 38.1 % MBO 10.8 % Secondary school 6.5 % Primary school 0 % Average number of restaurant visits per month 0 times 12.9 %

1 – 2 times 66.2 %

3 – 4 times 15.1 %

More than 4 times 5.8 % Average amount spend per person per restaurant visit 0 – 15 euro 4.3 % 16 – 30 euro 48.2 % 31 – 45 euro 38.8 % More than 45 euro 8.6 % Familiarity with consumer review systems Mean 4.93

SD 1.40

Min 1

Max 7

Familiarity with professional review systems Mean 3.27

SD 1.47

Min 1

Max 6.5

In order to check whether the characteristics of the respondents in the experimental were similar compared to the control groups several independent t-tests and a chi-square test were conducted. The original SPSS output can be found in Appendix 2.

As can be seen from table 3, an independent samples t-test showed no significant difference between the experimental group and the two control groups with regard to age (t (136) = 1.23, p = .221), familiarity with consumer review systems (t (137) = .61, p = .546) and familiarity with professional review systems (t (137) = .54, p = .591).

In addition, as can be seen from table 4, a Chi‐square test showed no significant relation between the experimental group and the control groups with regard to gender (χ² (1) = .53, p = .468), educational level (χ² (3) = 4.77, p = .189), average number of restaurant visits per month (χ² (3) = .38, p = .945), and average amount spend per person per restaurant visit (χ² (3) = 2.95, p = .399).

(20)

20

Table 3. Difference between the experimental and two control groups with regard to age, familiarity with consumer review systems and familiarity with professional review systems

Variable Control group

Experimental group (status)

N M SD N M SD Sign. (p)

Age 71 36.75 15.71 68 33.66 15.69 .221

Familiarity with consumer review systems 71 4.87 1.37 68 5.00 1.45 .546 Familiarity with professional review systems 71 3.16 1.45 68 3.38 1.51 .591 ***p <.001; **p <.01; *p<.05;°p<.10

Table 4. Difference between the experimental and two control groups with regard to gender, educational level, average number of restaurant visits per month and average amount spend per person per restaurant visit

Variable Chi-square DF Sign. (p)

Gender .53 1 .468

Educational level 4.77 3 .189

Average number of restaurant visits per month

.38 3 .945

Average amount spend per person per restaurant visit

2.95 3 .399

***p <.001; **p <.01; *p<.05;°p<.10

3.6 Experiment

3.6.1 Status motives

The status-related text and the non-status-related text, which were used in past research (Griskevicius et al., 2009, Griskevicius et al., 2010, Knechel & Leiby, 2016) to manipulate status motives, were translated to Dutch and checked by a native speaker of Dutch and English. In the status-related text, respondents had to imagine that they just graduated and started working at a prestigious company. In the story, respondents were motivated to attain a higher status within the company. At the end of the story respondents learned that they are competing with their peers for an important promotion. In the story, status-related elements, such as the luxury cars at the parking lot, were emphasised to manipulate status motives. In the non-status related story, respondents had to imagine that they lost their ticket for an upcoming concert. They searched for it throughout their home. At the end of the story, respondents were told that they found the ticket and they go to the concert with a peer.

(21)

21

3.6.2 Manipulation check status motives

A manipulation check was conducted to check whether the manipulation of status motives was successful and if the scripts elicited similar levels of affect. In addition, it was examined whether the two texts scored similar on narrative transportation. Narrative transportation in this study is defined as ‘a convergent process, where all mental systems and capacities become focused on

events occurring in the narrative’ (Green & Brock, 2000, p. 701) The narrative transportation

scale was used to measure to what extent the respondents were transported into the world of the narrative. The two texts should score high on narrative transportation since respondents can only be manipulated by the text if they really were involved with the story and the main characters. The manipulation check scales can be found in table 7 and the Cronbach’s Alfa’s of the scales in paragraph 3.6.6.

In the pre-test, the manipulation of status motives was not significantly different for the status-related and non-status-related text (t6) = .91, p = .396). The means were, however, in the right direction (status = 5.3, control = 4). To make sure that the manipulation of status motives would be successful in the real experiment, a few things were adjusted.

The manipulation stories were originally from a study conducted in the United-States (Griskevicius et al., 2009). The story was translated to Dutch, but since it was a direct translation, some things that in the United-States are related to high status, such as antique furniture, might not be considered as high status in the Netherlands. Therefore, the texts were adjusted for the real experiment to make them more applicable to a Dutch context.

In addition, respondents in the pre-test did not know if they had to answer the status manipulation text questions based on the feelings that they had after reading the story or based on their own personality. Therefore, a sentence was added that asked the respondents to keep trying to feel the emotions of the main character and fill in the questions based on these emotions.

Finally, to make sure that the respondents read the text carefully, in the real experiment it was added that the study consists of two parts, namely a memory recall task and questions about online reviews. At the end of the experiment, a question about the text was asked to check whether the respondents read the text carefully.

After these adjustments based on the results of the pre-test were made, the real experiment was conducted. In the real experiment, as can be seen from table 5, an independent samples t-test showed a significant difference between the experimental group and control group with regard status motives (t (99) = 2.26, p = .026). Respondents who read the status-related text (M = 5.01, SD = 1.76) were shown to score higher on status motives than

(22)

22

respondents who read the non-status-related text (M = 4.21, SD = 1.46). Since the status-related text also scored higher on status motives than the non-status-related text, the manipulation of status motives was successful and the two texts could be used to manipulate status motives.

In addition, as can be seen from table 5, an independent samples t-test showed no significant difference between the experimental group and the control group with regard to positive affect (t (99) = .85, p = .465) and negative affect (t (99) = .42, p = .399). Therefore, the non-status-related text and the status-related text elicited similar levels of affect which is good, because in this way affect elicited by a particular text did not influence user’s choice for an online review system.

Finally, as can be seen from table 5, an independent samples t-test showed a significant difference between the experimental group and control group with regard to narrative transportation (t (99) = 3.25, p = .002). Although the narrative transportation of the non-status-related story (M = 6.39, SD = 1.04) was higher than the status-non-status-related story (M =5.66, SD = 1.07), the status-related story still scored high on narrative transportation. A high score on narrative transportation is important, because respondents can only be manipulated by the text if they are transported into the world of the narrative.

Table 5. Differences of status motives, affect and narrative transportation between the status-related and the non-status-related text.

Variable Status-related text Non-status-related text

N M SD N M SD Sign. (p) Status motives 68 5.01 1.76 33 4.21 1.46 .026* Positive affect 68 1.38 5.24 33 1.28 5.48 .465 Negative affect 68 1.45 3.22 33 1.48 3.09 .399 Narrative transportation 68 5.66 1.07 33 6.39 1.04 .002** ***p <.001; **p <.01; *p<.05;°p<.10

3.6.3 User’s choice for an online review system

Firstly, respondents had to imagine if they had to make a reservation for a restaurant in a country where they had never been before and were not familiar with the review systems. In the pre-test, participants were asked to imagine if they were on holiday. Some participants mentioned that they are less critical when they are on holiday and therefore they would choose a consumer review system. Since it was found that the particular holiday situation influenced the choice of some participants for a certain system, participants in the real experiment were asked to imagine if they were in a country where they never been before. This situation was chosen since in this

(23)

23

way it would be realistic that the respondents were not familiar with the review systems in the country. It was considered to be important to not work with existing review systems, such as IENS and Lekker, because brand familiarity and brand attitude could have biased the answers of the respondents. Brand familiarity is found to reduce mental effort when making purchase decisions (Laroche, Kim & Zhou, 1996). When a respondent is, for example, familiar with IENS but not with Lekker, this respondent might be more inclined to choose for IENS, only because he or she is more familiar with the brand and not because of a preference for a certain type of online review system.

Secondly, respondents were told that they read about two popular online review systems on a Dutch website about the country and information about both systems was given. In the pre-test, the information about both systems was presented on website screens. Although the researcher tried to make the websites of both systems as similar as possible, several respondents still mentioned that they favoured one website design over the other. To avoid that website design would influence user’s choice for an online review system, it was decided to only present textual information about the review systems to the respondents in the real experiment. In this way, respondents could really base their decision on differences between the systems instead of differences in design of the two systems.

For the professional review system, it was explained that the reviews on the system are written by culinary journalists who have a background in the food service industry or are wine professionals and judge the restaurants on 60 criteria. For the consumer review system, it was explained that the reviews are written by other consumers who already visited the restaurant. In addition, it was mentioned that you can also write reviews yourself on the consumer review

system.

The information texts about the two review systems in the real experiment were similar to the texts in the pre-test since all the participants in the pre-test could correctly explain the difference between the two systems. However, some adjustments to the original text were made based on the thoughts and answers of the participants in the pre-test. In the pre-test, participants mentioned that they favoured the consumer review system, because they thought the professional review system would only contain expensive restaurant and no restaurants that they could afford. Since in reality, professional review systems do contain different types of restaurants for all budgets, the text was adjusted and for both systems it was mentioned that all restaurants of the country could be found in the system. Additionally, the sentence that consumers could only have access to the consumer review system, if they would register

(24)

24

themselves was removed since this could be a barrier to use the system and as a result influence user’s choice for an online review system.

Thirdly, respondents had to decide which review system they would choose: Dine Out, a professional review system or Restaurant World, a consumer review system. The names Dine Out and Restaurant World were given to the systems to make the situation more realistic which increased the ecological validity of the study (Field, 2009). English names were chosen since respondents had to imagine if they were in a foreign country and these particular names were chosen because they are simple and they are considered to be appropriate for the product category restaurants.

In order to control for order effects (Field, 2009), the answer categories (professional review system vs. consumer review system) of the question ‘which review system would you choose?’ and the explanations of the review systems in the text were shown in a random order to respondents. In addition, to gain more insight into the motivation of users to choose a professional or consumer review system, a qualitative question was asked, namely ‘Why did you chose Dine Out?’ or ‘Why did you chose Restaurant World?’.

.

3.6.4 User’s expertise

After respondents chose the online review system, the expertise of the subject with regard to restaurants was measured. As can be seen from table 7, to measure user’s expertise the domain specific consumer knowledge scale of Luo and Toubia (2015) was used. Luo and Toubia (2015) used this scale to measure the knowledge for consumers in the domain of fast-food restaurants, personal banking, movie theatres and social media platforms. For this study, the domain-specific consumer knowledge scale was adapted to the restaurant domain. The domain-domain-specific consumer knowledge scale consisted of seven five-point Likert scales. Item 1 and 3 were negatively formulated and therefore reverse coded. The items were negatively formulated in order to avoid the response set phenomenon, which is the tendency of participants to give a similar response to a set of items, regardless of the meaning of the items (Maassen, 1991). A remedy for this is to negatively formulate some items in a scale, because this makes participants more alert when they fill in a questionnaire.

(25)

25

3.6.5 Control variables

As can be seen from table 7, eight control variables were included, namely age, education, gender, familiarity with a professional review system, familiarity with a consumer review systems, number of restaurant visits per month, average amount spent per restaurant visit per person and purpose of the study.

3.6.6 Contruct reliability and validity

Convergent validity can be defined as ‘how well subscales correlate with other measures that

are assumed to be related’ (Meirte et al., 2016, p. 86). To access the convergent validity of the constructs, several principal component factor analysis were conducted. There are several assumptions which should be met to be able to conduct a factor analysis (Field, 2009). Firstly, the Kaiser-Meyer-Olkin (KMO) should be above .5. Secondly, Bartlett’s Test should be significant and therefore less than .5. Lastly, a criterium in social science is that a factor solution should at least account for 60 percent of the total variance (Hair et al., 2014). In addition, Cronbach’s Alpha’s were calculated for each scale since Cronbach’s Alpha is a good measure to determine the reliability of a questionnaire to measure a certain construct (Field, 2009). A Cronbach’s Alpha of .70 or higher is an indication that items consistently reflect the same construct. The original SPSS output can be found in Appendix 2.

As can be seen in table 6, the reliability of status motives comprising two items was good: α = .94. Therefore, a compound variable for status motives was created. In addition, a principal component analysis revealed a one factor solution, explaining 94 percent of the variance (KMO = .50, Bartlett <.001).

The reliability of positive affect after reading the stimulus text comprising two items was acceptable: α = .78. Therefore, a compound variable was created. A principal component analysis revealed a one factor solution, explaining 83 percent of the variance (KMO =.50, Bartlett <.001). The reliability of negative affect after reading the stimulus text comprising two items was acceptable: α = .74. Therefore, a compound variable was created. A principal component analysis revealed a one factor solution, explaining 80 percent of the variance (KMO =.50, Bartlett <.001).

The reliability of narrative transportation comprising five items was acceptable: α = .71. A principal component factor analysis revealed that a one factor solution explained 48 percent of the variance (KMO = .68, Bartlett <.001). This is, lower than the criteria of 60 percent explained variance (Hair et al., 2014). However, when items were deleted the percentage of

(26)

26

explained variance still not met the criteria of 60 percent. Therefore, it was decided to use the original five item scale of Green and Donahue (2008).

The reliability of user’s expertise comprising seven items was good: α = .89. Therefore, a compound variable for user’s expertise was created. A principal component analysis revealed a one factor solution, explaining 62 percent of the variance.

The reliability of familiarity with a professional review system comprising two items was acceptable: α = .72 and the reliability of familiarity with a consumer review system comprising two items was good: α = .82. Therefore, compound variables were created. A principal component analysis for familiarity with professional review systems revealed a one factor solution, explaining 78 percent of the variance (KMO = .50, Bartlett <.001). In addition, a principal component analysis for familiarity with consumer review systems revealed a one factor solution, explaining 85 percent of the variance (KMO = .50, Bartlett <.001).

Table 6. Internal consistency and convergent validity

Construct Original nr of items

Cronbach’s alpha Original nr of items deleted

Cronbach’s alpha Percentage explained variance

Manipulation check status texts

Status motives 2 .94 0 94 % Positive affect 2 .78 0 83 % Negative affect 2 .74 0 80 % Narrative transportation 5 .71 0 48 % Independent variable User’s expertise 7 .89 0 62 % Control variables Familiarity with a professional review system 2 .72 0 78 % Familiarity with a consumer review system 2 .82 0 85 %

(27)

27

3.7 Interview

After the experiment was conducted, a restaurant owner, namely Leander Kattestaart, was interviewed. This interview was conducted to discuss the results of the current study with him and to look at user’s choice for an online review system not only from a user’s perspective but also from a restaurant owner perspective. Leander Kattestaart is owner of restaurant L’Ambiance in Middelharnis. Restaurant L’Ambiance is a relatively expensive restaurant, which serves dishes inspired by the French kitchen made with local and fresh ingredients. The prices range from 39.50 euro for a three-course dinner to 63.50 euro for a six-course dinner (Restaurant L’Ambiance, 2016).

(28)

28

Table 7. Operationalization variables

Construct Definition Scale Items

Independent variables

User’s expertise (Luo & Toubia, 2015)

A person’s embodied skill or expertise that is acquired

through past experiences, formal or informal training, or

education (Luo & Toubia, 2015,

p.102)

• Seven seven-point Likert scales were used (Luo and Toubia (2015) used five-point Likert scales)

Compared to the average person

1. I do not know much about restaurants (reverse-coded) 2. I am very familiar with restaurants

3. I am not knowledgeable about restaurants (reverse coded)

4. I am very interested in restaurants 5. I go to restaurants a lot

6. My friends go to restaurants a lot

7. I read about restaurants (e.g., reviews, blogs, inserts, ads, flyers) all the time

Manipulation check

Status motives (Griskevicius et al., 2009)

‘The desire to gain prestige, respect, or power over others’ (Griskevicius

et al., 2009).

• Two seven-point Likert scales were used (Griskevicius et al. (2009) used nine-point Likert scales)

To what extent:

1. Do you desire to have higher social status 2. Are you motivated to have higher prestige

Narrative transportation

‘a convergent

process, where all mental systems and capacities become

focused on events occurring in the narrative’ (Green & Brock, 2000, p.

701)

• Five seven-pont Likert scales were used (Green & Donahue, 2008).

Indicate to what extent the following statements describe you:

1. I could picture myself in the scene of the events described in the narrative.

2. I was mentally involved in the narrative while reading it.

3. I wanted to learn how the narrative ended 4. The narrative affected me emotionally.

5. While reading the narrative I had a vivid image of the main character

(29)

29

Construct Definition Scale Items

Manipulation check

Positive affect towards the manipulation text

• Two seven-point Likert scales were used (Griskevicius et al. (2009) used nine-point Likert scales)

To what extent…

1. Do you feel enthusiastic? 2. Do you feel excited? Negative affect towards

the manipulation text

• Two seven-point Likert scales were used (Griskevicius et al. (2009) used nine-point Likert scales)

To what extent…

1. Do you feel frustrated? 2. Do you feel angry?

Dependent variables

User’s choice for online review system

The choice for an online consumer review system or an online professional review system.

• Item 1: 2 answer categories (The answer categories were shown in a random order)

• Item 2: open question

Which system do you choose? 1. Dine Out

2. Restaurant World

Why did you chose for Dine out/Restaurant world?

(30)

30 Construct Definition Scale Items

Control variables

Age What is your age?

Gender What is your gender?

1. Man 2. Woman Purpose of the

study

What is the purpose of the study according to you?

Education Two questions 1.What is the highest educational level that you have completed? 1. Primary school

2. Secondary school 3. MBO

4. HBO 5. WO

2. What did you study?  will only be shown if a participant filled in MBO, HBO or WO to check whether someone did a study in hospitality management

Familiarity with professional review systems

Two questions Indicate what your opinion is about the following statements. 1. I am familiar with professional review systems

2. I use professional review systems a lot Familiarity with

consumer review systems

Two questions Indicate what your opinion is about the following statements. 1. I am familiar with consumer review systems

2. I use consumer review systems a lot

Usage experience 1. Indicate how often you go to a restaurant on average per month. 1. 0 times

2. 1 -2 times 3. 3 – 4 times 4. More than 4 times

2. Indicate how much you spend on average per person when you go to a restaurant. 1. 0 – 15 euro

2. 16 – 30 euro 3. 31 – 45 euro 4. More than 45 euro

(31)

31

4. Results

4.1 Descriptive analysis

A correlation analysis was conducted with the independent variable expertise and the control variables that were measured on an interval measurement level. The output of the original SPSS analysis can be found in Appendix 3.

As can be seen from table 8, expertise correlated highly with familiarity with professional review systems and familiarity with consumer review systems. The more expertise a respondent had, the more familiar the respondent was with both professional and consumer review systems. In addition, familiarity with professional review systems correlated with familiarity with consumer review systems. The more familiar a respondent was with professional review systems, the more familiar the respondent was with consumer review systems. Furthermore, age correlated negatively with familiarity with consumer review systems. The younger a respondent was, the more familiar the respondent was with consumer review systems. Finally, age correlated negatively with expertise. The younger a respondent was, the more expertise about restaurants the respondent had.

Table 8. Correlation matrix and descriptive statistics

Expertise Familiarity with professional review systems Familiarity with consumer review systems Age Expertise

Familiarity with professional

review systems .425***

Familiarity with consumer

review systems .379*** .291**

Age -0.146° -0.078 -.277**

Mean 4.23 3.27 4.93 35.24

SD 1.18 1.47 1.40 15.72

n = 139; *** p<.001; **p<.01; *p<..05; °p<.10

The dependent variable user’s choice for an online review system was also measured. In total, 110 respondents chose the consumers review system and 29 chose the professional review system.

In order to control for any particular aspect of the control story, the current study included a second control condition in which participants did not read any story. Similarly to the procedure of the study of Griskevicius et al. (2010), it was predicted that the two control conditions would not differ with regard to user’s choice for an online review

(32)

32

system. As can be seen in table 9, in line with this prediction, a binary logistic regression revealed no significant difference between the two control groups with regard to user’s choice for an online review system (p = .714). As a result, the two control conditions were combined for further analysis. The SPPS output can be found in Appendix 2.

Table 9. Binary logistic regression analysis to examine the effect of the control groups to predict user’s choice for an online review system

B S.E. Wald df Sig. Exp(B)

Control groups -0.24 0.64 0.13 1 .714 0.79 Constant -1.49 0.42 12.65 1 .000*** .023 n = 71; *** p<.001; **p<.01; *p<..05; °p<.10

4.2 Assumptions

The assumptions for a binary logistic regression are: independent variables of interval measurement level, dichotomous dependent variable, linearity, normality and independence of observations (Field, 2009).

Firstly, dummy variables were created for the categorical variables education, average number of restaurant visits per month and average amount spend per restaurant per person. In this way, all the independent variables were of interval measurement level.

Secondly, the dependent variable user’s choice for an online review system is a dichotomous variable, namely the choice between a professional online review system and a consumer online review system, and therefore binary logistic regression is appropriate.

Thirdly, an assumption is linearity between continuous predictors and the logit of the outcome variable (Field, 2009). This assumption is tested by conducting a binary logistic regression and looking at the significance level of a relationship between an independent variable and the outcome variable.

Fourthly, normality could be assumed since the sample size for all groups w as at least 30 which is necessary to assume normal distribution (Hair et al., 2014). In addition, the skewness and kurtosis values for the variables were between the -3 and +3 which is also an indication of a normal distribution.

Fifthly, the observations were independent since respondents were randomly assigned to one of the conditions. Therefore, there was no connection between the observations in the different groups.

(33)

33

4.3 Hypothesis testing

A binary logistic regression analysis was conducted to predict user’s choice for an online review system using user’s expertise, status motives, age, gender, education, familiarity with consumer review systems, familiarity with professional review systems, average number of restaurant visits per month, and average amount spend per person per restaurant visit as predictors. Firstly, a binary logistic regression analysis was conducted with only the control variables. A test of the full model against a constant model was statistically significant, indicating that the predictors as a set reliably distinguished between user’s choice for a professional review system or a consumer review system (chi square = 31.147, p = .003 with df = 13). The Nagelkerke R² was .313 indicating that the total model explained 31.3 percent of the variance in user’s choice for an online review system. Prediction success was 79.9 percent (92.7 percent for the consumer review system, 31 percent for the professional review system).

As can be seen in table 10, the Wald criterion demonstrated a significant contribution of familiarity with consumer review systems (β = -.58, p = .008) and familiarity with professional review systems (β = .54, p = .007) to predict user’s choice for an online review system. In addition, average number of restaurant visits per month dummy 1 (β = -2.14, p = .090), average amount spend per person per restaurant visit dummy 2 (β = -1.68, p = .048) and average amount spend per person per restaurant visit dummy 3 (β = -1.37, p = .091) were found to significantly contribute to predict user’s choice for an online review system. Secondly, a binary logistic regression analysis was conducted with both the independent variables, user’s expertise and status motives, and the control variables. A test of the full model against a constant only model was statistically significant, indicating that the predictors as a set reliably distinguished between user’s choice for a professional review system or a consumer review system (chi square = 38.954, p = .001 with df = 15). The Nagelkerke R² was .381 indicating that the total model explained 38.1 percent of the variance in user’s choice for an online review system. Prediction success was 81.3 percent (93.6 percent for the consumer review system, 34.5percent for the professional review system). Compared to the model where only the control variables were included, the percentage of total variance explained increased with 6.8 percent.

As can be seen in table 11, the Wald criterion demonstrated a significant contribution of expertise (β = .86, p = .010) to predict user’s choice for an online review system. The odds ratio was 2.36, indicating that respondents were 2.36 times as more likely to choose for a professional review system when their expertise is raised by one unit. As a result, hypothesis 1

(34)

34

was confirmed. This study revealed that if the expertise of the user is high (low), then the user is more (less) inclined to choose an online review system based on professional reviews.

In contrast, hypothesis 2, which predicted that a user (not) primed with status motives is (less) more inclined to choose a professional review system, was not confirmed. Status motives (β = .30, p = .577) did not make a significant contribution to predict user’s choice for an online review system.

Similar to the model were only the control variables were included, familiarity with a consumer review system (β = -.70, p = .004) and familiarity with a professional review system (β = .38, p = .073) significantly contributed to predict user’s choice for an online review system. The odds ratio for familiarity with a consumer review system was .50, which indicates that respondents were .50 times less likely to choose for a professional review system when their familiarity with a consumer review system was raised by one unit. The odds ratio for familiarity with a professional review system was 1.46, which means that consumers were 1.46 times more likely to choose a professional review system when their familiarity with a professional review system was raised by one unit.

When user’s expertise and status motives were added to the model, average number of restaurant visits per month dummy 1 (0 vs. >4) and average amount spend per person per restaurant visit dummy 2 (16-30 vs. >45) and dummy 3 (31-45 vs. >45) did not make a significant contribution to predict user’s choice for an online review system anymore.

What to Choose? A Study on User’s Choice for an Online Review System.