The differences between reviews on Facebook and online review platforms : a link between online writing motivations and sentiment

(1)

The differences between reviews on Facebook

and online review platforms

A link between online writing motivations and sentiment

Berend Bom (10002301) University of Amsterdam

Master Business Administration – Digital Business

Master Thesis (final version)

Supervisor:

Abhishek Nayak June 23, 2017

(2)

II

Statement of originality

This document is written by Student Berend Bom who declares to take full

responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and

that no sources other than those mentioned in the text and its references have

been used in creating it.

The Faculty of Economics and Business is responsible solely for the

(3)

III

Abstract

Consumers rely more on written reviews (user-generated-content) than on the

summary of a product of service. Therefore, many websites and social media provide the

opportunity to write a review. This study focuses on reviews in the restaurant industry and

makes a distinction between reviews on Facebook and reviews on online review platforms

like Yelp and TripAdvisor. Furthermore, it makes a distinction between platforms and delves

deeper into the differences in writing motivation and sentiment between their respective users.

The study is founded upon two different datasets. The first dataset (N = 217) is obtained

through a questionnaire among Dutch citizens that inquires after their main motivations for

writing online restaurant reviews. The second dataset is a collection of online restaurant

reviews that consumers posted on Facebook and Yelp (N = 1588). The main results

concerning Facebook suggest that the most important writing motivations are concern for

others, negative feelings, helping the company, and positive self enhancement. For online review platforms, the most important motivations for writing a review are negative feelings

and social aspects. The analysis of the review database shows that reviews are in general

more positive on Facebook. The reviews on Facebook are less nuanced compared to online

review platform reviews, meaning that positive Facebook reviews are more positive than the

positive reviews on online review platforms. Likewise, negative reviews on Facebook are

more negative. This study will aid in advancing knowledge about restaurant reviews,

shedding light on what motivates reviewers to post reviews online, but also on the differences

in sentiment between the online platforms.

Keywords: writing motivations, Facebook, online review platforms, sentiment, word of

(4)

IV

Acknowledgements

I would like to thank Dr. Abhishek Nayak for his involvement and the feedback he gave me

during this master thesis process. Furthermore, I would like to thank Joachim Buur, Emma

Bokwinkel and Holger Jansen in particular, but also my fellow students and roommates for

(5)

V

List of figures and tables

Figures:

Figure 1. Social media use worldwide (Statista, 2017) ... 17

Figure 2: Conceptual framework ... 22

Tables: Table 1: Collected variables of the questionnaire ... 25

Table 2: Variables of the review database ... 31

Table 3: Frequencies of written reviews ... 34

Table 4: Descriptive statistics - control variables ... 34

Table 5: Descriptive statistics - writing motivations... 35

Table 6: Correlation table of Facebook (FB) variables and Cronbach's alpha score ... 37

Table 7: Correlation table of TripAdvisor (TA) variables and Cronbach's alpha score ... 37

Table 8: Logistic regression of positive intention and Facebook ... 39

Table 9: Logistic regression of negative intention and Facebook ... 40

Table 10: Logistic regression of positive intention and TripAdvisor ... 42

Table 11: Logistic regression of negative intention and TripAdvisor... 43

Table 12: Gender statistics... 45

Table 13: Gender differences ... 45

Table 14: Platform differences between sentiment score and stars ... 46

Table 15: Comparison between positive and negative intention ... 47

Table 16: Mixed intention compared with positive and negative intention ... 47

Table 17: Sentiment of intention related to Yelp and Facebook ... 48

Table 18: Differences between intention and platform ... 48

(6)

VI

1. Introduction

Last May, The Netherlands Authority for Consumers and Markets (ACM) published a

press release in which they announced that they will start enforcing rules and regulations to

govern online reviews. Online reviews are ratings bestowed by the consumer upon a product or

service. Because reviews are playing an increasingly prominent role in influencing the purchase

decisions of consumers, it is very important that the quality of these reviews is ensured.

Purchase motivation can be influenced by the quality and quantity of a review (ACM, 2017;

Park, Lee, & Han, 2007). According to Chevalier and Mayzlin (2006), people rely more on

written reviews (and other user-generated content) than on the summary of a product of service.

Therefore, many websites and social media platforms allow users to write reviews. Due to the

increasing importance of online reviews, the ACM wants more transparency in how the reviews

are established, and pushes for all parties who post comments online or allow others to post

comments to grant more insight into their review processes. Online reviews in the Netherlands

are subject to certain rules. For instance, consumers have the right to know how reviews are

collected. Furthermore, review platforms are required to check the relevance and quality of the

reviews, to employ systems which could detect fake reviews, and to be transparent about the

lead-up to a particular review. More specifically, consumers need to know if reviews are part

of a marketing strategy, and whether a reviewer received free products or other rewards as

compensation for writing review. Finally, the review web page needs to disclose any changes

made to a review after contact with the reviewer (ACM, 2017).

Online reviews are a relatively new phenomenon in marketing. Since the evolution of

the internet to Web 2.0 in 2005, traditional word of mouth marketing has evolved into electronic

word of mouth. Web 2.0 made it possible for users to interact online and create online content

on a large scale (Constantinides & Fountain, 2008). Since the rise of Web 2.0 different

(9)

2 evaluate and interact with each other in the context of their relationships with companies and

individuals. All their reviews and opinions affect other people, particularly when it comes to

information search and decision making (Herrero, San Martín, & Hernández, 2015). Because

user-generated content is helping, for example, restaurants improve their search result ranking

and attract new customers, the ACM will start guiding, advising and informing these companies

on how to comply with the review regulations (ACM, 2017). In May 2017, ACM conducted

their own study on online reviews. They concluded that in the future, online reviews will pose

a significant risk of deceiving consumers and harming other companies through misleading

review content.

The restaurant industry is one of the sectors in which customer reviews are most

common, and therefore one of the sectors which is most dependent on them (ACM, 2017;

Zhang, Ye, Law, & Li, 2010). And as the restaurant market is making increasing use of online

reviews, it is also growing in size. In 2015 the Dutch spent 643 million euros more on eating

out than the year before. According to the director of FoodService Instituut Nederland,

Jan-Willem Grievink (2015), this higher expenditure is due to growing consumer confidence. This

growth is still in progress (Dongen, 2016); in 2016, the revenues of the hospitality sector grew

by 8.8 per cent, with the revenues of restaurants and cafés growing most rapidly at a respective

3.5 and 3.4 per cent per quarter (ANP, 2016; CBS, 2016). The restaurant industry in the

Netherlands is growing year by year. In 2016 the Netherlands had over 11,000 restaurants, more

than 10,000 fast food companies, 2000 cafés and 1500 hotel restaurants. Furthermore,

star-awarded restaurants have also been growing in number over the last years (Rabobank, 2016).

Concurrent with this development, going out for dinner is gaining popularity in the Netherlands.

Compared with 2015, revenues gained from consumers going out for dinner have grown by

4.5% in 2017. The popularity of eating out is growing particularly fast in the age group of 18

(10)

3 is doing in real life. All over the Netherlands, trendy new restaurant concepts are opened

successfully (Schutijser, 2017).

Considering the increasing importance of online reviews, it would be interesting to

uncover more about the mechanisms of posting reviews. There are many platforms on which to

post restaurant reviews. Of these, the largest restaurant review platform is Yelp (Luca, 2016).

Alongside restaurants there is also the opportunity to review shops, nightlife venues, ‘active

life’ venues, beauty & spas, automotive companies, et cetera. TripAdvisor is the largest

platform pertaining to hospitality reviews. TripAdvisor looks similar to Yelp, though it boasts

fewer restaurant reviews and more travel information. Indeed, they are the most prominent

travel platform (Gretzel & Yoo, 2008). TripAdvisor also owns Iens.nl, a review platform for

restaurants in the Netherlands, and there are smaller competitors such as OpenTable. In short,

there is ample opportunity for consumers to review the restaurants they have visited, and

consumers have plenty of choice between review providers.

Next to these review platforms there is social media. Social media platforms, like

Facebook, MySpace and YouTube, allow users to create a personal profile and share

information with others. Among other things, users can share and follow information about

companies or products (Ross et al., 2009). Moreover, since the summer of 2015 users can post

reviews about companies on Facebook, including restaurants. This makes it possible to not only share one’s favourite dinner spot with friends, but also to look at reviews and find out what others think about the restaurant (Mashable, 2015). As such, users can turn to Facebook

alongside regular review platforms to review and read reviews. Because Facebook constitutes

an entirely new type of review platform, with a lot more users than regular review platforms,

restaurants may be interested in knowing the differences between the types. It is relevant for

restaurants to know if there is a difference between the reviews on different platforms. Is there

(11)

4 is good practice to know how online reviewers behave on the internet and consequently to know

how to interact with them or think of new marketing strategies (Daugherty, Eastin, & Bright,

2008). The present study will delve deeper into these differences between platform types.

Reviews are essentially a form of word of mouth. Much research has already been done

on Word of Mouth (WOM). Dichter (1966) addressed offline WOM; the effects of WOM on

the behaviour of consumers. Electronic WOM (eWOM) and its influence has been elaborately

researched since 2001 (Balasubramanian & Mahajan, 2001; Bickart & Schindler, 2001).

However, this past research has mostly focused on the ‘receivers’ (readers of online reviews)

of electronic word of mouth (eWOM) — in other words, about the impact of eWOM on

(purchase) behaviour (Gruen, Osmonbekov, & Czaplewski, 2006; Zhang, Ye, Law, & Li,

2010). Writing motivations for ‘senders’ have been much less explored (Hennig-Thurau,

Gwinner, Walsh, & Gremler, 2004). Hennig-Thurau (2004) was the first to discuss eWOM in

the context of users’ motivations for generating it. Hennig-Thurau’s study is one of the few

which has thus far addressed online writing motivations. However, the little research that has

been done has focused on online reviews as a whole, without distinguishing between platforms.

Yet, in the case of the restaurant industry, the review options for customers are extensive and

differ in terms of length, type and accessibility of the reviews. As such, reviews and their

motivations are likely to vary between platforms. Additionally, no research has been done on

Facebook reviews. In order to be able to properly tailor their strategies to each separate review

platform, restaurants will have to know how consumers’ writing motivations differ between

platforms. This study aims to fill that gap in knowledge.

In a casual, exploratory comparison between different platforms suggests that on

Facebook the reviews seem more positive and shorter in contrast to regular review platforms.

(12)

5 contrast with Yelp reviews. This research will delve deeper into the differences in review

sentiment between the platforms and investigate whether there are differences between writing

motivations. What motivates someone to post a positive or negative review on Facebook rather

than on a regular online review platform? And do users tend to be more positive or negative on

Facebook? In other words, do the platforms differ in terms of average review sentiment? The

answers to these questions will be the main contribution of this study.

In summary, this research attempts to fill an extensive gap in knowledge about online

reviews. It makes a distinction between platforms and delves deeper into the writing

motivations and sentiment differences of reviews on Facebook and online review platforms.

The central research question that is answered is:

To what extent is there a difference between online review platforms and social media (Facebook) in terms of writing motivations and sentiment of online reviews?

(13)

6

2. Literature Review

This section addresses the theoretical background of the present study. The first part

discusses the definition of Word of Mouth (WOM) and electronic Word of Mouth (eWOM).

The second part explores the effectiveness of eWOM. The third part examines consumers’

motivations for writing online reviews. The fourth part describes the external factors — besides

motivations — which may play a role in shaping the sentiment of a review (how positive or

negative a review is). The fifth part details the differences between social media and online

review platforms with regards to the reviews published on them. Finally, the findings of the

literature review are used to formulate hypotheses, which are then combined into a conceptual

model.

2.1 WOM and eWOM

In the context of commerce, WOM is face-to-face communication between people

before any purchase has been made. Potentially functioning as a type of marketing, this form

of communication takes place between buyers and non-buyers in an informal setting (Arndt,

1967). Harrison-Walker (2001) uses a similar definition: “WOM may be defined as informal,

person-to-person communication between a perceived noncommercial communicator and a receiver regarding a brand, a product, an organization, or a service”. These two definitions, written decades apart from each other, are nearly the same, which could mean the definition of

WOM is fixed (Jeong & Jang, 2011).

Nowadays, with Web 2.0, there is a plethora of opinion platforms where customers can

share their opinions and experiences concerning both products/goods and services. A new

phenomenon of eWOM is “viral marketing”, which describes an exponential growth in (positive or negative) popularity of a brand as a result of consumers who are spreading the

(14)

7 other things, relationships between organizations and customers. The definition of eWOM is

“any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet” (Hennig-Thurau et al., 2004; Stauss, 2000).

The main difference between WOM and eWOM is found in the relationship between

the communicator and receiver. In contrast to WOM, recommendations on the internet

(eWOM) come mainly from people unknown to the receiver and are primarily text-based,

which makes it more difficult to draw the correct conclusions about the credibility of the

message (Chatterjee, 2001). On the other hand, this anonymity is also an advantage; it facilitates

self-exposure, and therefore generally makes it easier to share experiences. In conclusion, “compared to traditional WOM, online WOM is more influential due to its speed, convenience, one-to-many reach, and its absence of face-to-face human pressure” (Sun, Youn, Wu, & Kuntaraporn, 2006).

Effectiveness of eWOM

In one of Kotler’s first marketing books, he already addressed the impact of (offline) WOM. He states that although the traditional way of advertising does indeed influence a person’s behaviour, it is nevertheless not as influential as influences from other associates and a consumer’s own observations (Kotler, 1967). Later research indeed claimed WOM has more impact on customers’ behaviour compared to other marketing strategies such as print

advertising or billboards (Herr, Kardes, & Kim, 1991). Herr et al. (1991) demonstrated through

experiments that WOM communications have more impact because of their relatively high

accessibility and intelligibility. WOM makes products and services easier to remember and

interpret. Indeed, according to Sheth (1971), WOM makes customers more aware of the

(15)

8 The same goes for eWOM, only in an online context. EWOM helps firms increase their

value in the long run, but also aids their marketing effort in drawing in more customers. With

eWOM it is easier to approach a multitude of consumers than it is using the old-fashioned way

(WOM), and consumers remember the company better. Villanueva, Yoo and Hanssens (2008)

investigated the differences in effectiveness between WOM, regular marketing and eWOM.

Participants registered with a web host company, and of the people who remembered the

company from traditional marketing, less than 30% stayed active clients and showed returning

patterns on the website. Of the eWOM customers, this was more than 50%. As such, customers

acquired through eWOM add almost twice as much value as customers who registered through

traditional sources (Villanueva, Yoo, & Hanssens, 2008).

There is also research that examined the effectiveness of different sentiments in eWOM

(online reviews). The effectiveness of positive eWOM sentiments and negative eWOM

sentiments has been compared. Liu (2006), for example, concluded that positive reviews

boosted sales. Conversely, the influence of negative reviews seems to be even stronger. When

Chevalier and Mayzlin (2006) compared the influence of positive and negative review

sentiments on book sales, they concluded that negative reviews (ranked one-star) had more

impact on sales than high-ranked reviews. The number of reviews, too, has an effect; the higher

the number of likes and reviews, the higher the sales will be on a specific website (Chevalier &

Mayzlin, 2006). This corresponds with research by Liu (2006), which investigated reviews of

movies that were collected from the Yahoo Movie Website. Liu (2006) concluded that an

increase of eWOM about a movie is associated with a growth in sales. Furthermore, people like

to speculate about a movie prior to its release. They already start reviewing the movie

pre-release, which creates awareness, which in its turn leads to more reviews; as such, “firms should

try to create active WOM communication among potential use” (Liu, 2006). The more awareness exists among consumers, the better the sales will be (Sheth, 1971).

(16)

9 Apart from the regular review platforms, eWOM also occurs on social media. Online

communities, mainly Facebook, grow every year in terms of active users. On these social media

platforms people share, like and review content. In this way they are generating online content

themselves. “In contrast to other Internet businesses, online communities rely on

user-generated content to retain users” (Trusov, Bucklin, & Pauwels, 2009). User-user-generated content (UGC) on Facebook, such as reviews, shares and likes, has a positive effect on the reputation

and equity of a brand. The influence of UGC is much stronger for firms who communicate over

social media. For the best marketing results, companies should take UGC communication into

account. The power of consumers is strong, so maintaining an active social media platform is

essential for optimally influencing customers. Moreover, social media has the additional benefit

of potentially generating a viral response, which consequently leads to the spread of

firm-created content to a wider public; this affects the customer’s attitude toward the brand

(Schivinski & Dabrowski, 2016).

2.2 Writing motivations for eWOM and UGC

Understanding what drives consumers to create UGC could yield powerful insights for

advertisers and marketers, which might help them make more effective use of WOM. It is

known that positive WOM has an effect on purchase behaviour, but that negative WOM has an

even stronger impact on this behaviour (Arndt, 1967). Sen and Lerman (2007) link this to the

online world. Negative reviews have more impact because a negative bias involves a more

critical and scrupulous treatment of the product or service, and consumers tend to trust negative

reviews more. Attention is captured first by negative reviews, which is why a reader engages

more with online negativity than online positivity. Review readers believe a negative review is

closer to the reviewer’s true experience. Knowledge about what motivates customers to review

(17)

10 motivations, harnessing eWOM to boost their sales through positive reviews while avoiding

the loss of clientele through negative ones (Sen & Lerman, 2007).

Cheung and Lee (2012) used a questionnaire to uncover the reasons for consumers to

review online. In other words, they studied the reasons for consumers to engage in

review-based eWOM. There are three main antecedents responsible for eWOM: (1) reputation, (2)

sense of belonging, and (3) enjoyment of helping. (1) Considerations of reputation may be reason for a consumer to want to look like an expert when posting a review. They might want

to contribute by sharing their experiences with a large group of (potential) consumers (Cheung

& Lee, 2012). A consumer’s sense of belonging refers to their emotional involvement with the

group they are addressing (Cheung & Lee, 2012). According to Hard and Ou (2001), this can

be related to Maslow’s Hierarchy of Needs (Maslow’s pyramid). It corresponds with the need

for love and belonging. Online, people adjust their goals to those of the group they wish to be

part of. They wish to be part of a wider community by identifying themselves in the way the

community identifies itself (Lakhani & Von Hippel, 2003). Writing motivations pertaining to

this desire to be part of, and participate, in a community, fall under the social aspects of what

drives consumers to write reviews. Reviewing on online platforms may lead to returning social

benefits from the community (Hennig-Thurau et al., 2004). For example, in the case of

Yelp.com, this can come in the form of recognition. The more reviews a consumer puts on the

website, the more well-known they will become on the website. The third antecedent,

enjoyment of helping, refers to helping a group or company. Consumers engage in eWOM because they want to give something back to a company in return for the positive experience

they have been given. They may feel like their opinion will help others in the future (Cheung

& Lee, 2012; Hennig-Thurau et al., 2004; Tong, Wang, & Teo, 2007). Another reason for ‘helping’ is the equity theory. When a consumer expects to receive some form of ‘return of investment’ — for example, a sense of satisfaction from being able to help others — that is

(18)

11 proportionally larger than their own contribution, they could start recommending what the firm

has to offer by way of positive eWOM. The consumer decides if it is worth their time to review

their experience with the product or service (Hennig-Thurau et al., 2004).

Gangadharbatla (2013) investigated what drives Facebook users to post comments,

share content, or like information. Their findings on users’ motivation for creating UGC on

Facebook are partly in line with those suggested by Daugherty et al. (2008). The social function

of Facebook helps consumers find affiliation and belonging with a group and make new friends.

It is concerned with the need to belong and the need for cognition and confidence boosts.

Consumers adapt to and are influenced by social networking sites like Facebook for four

reasons: “Internet self-efficacy, need for cognition, need to belong, and collective self esteem”

(Gangadharbatla, 2008). This is a form of positive self-enhancement (Hennig-Thurau et al.,

2004).

Besides positive reviews, consumers may also post negative comments. In recent years

consumers have become more aware of their power online. With Web 2.0 and eWOM the power

has shifted from company to consumer. Consumers wield this power by way of reviews,

especially when they want to share criticism of a firm. On Web 2.0 there are a lot of potential

readers of reviews. Bad reviews operate as bad marketing for the company (Hennig-Thurau et

al., 2004). The motivation to criticize is a form of sharing negative feelings.

According to Hennig-Thurau et al. (2004), the antecedents for eWOM mentioned earlier

are only part of a larger package. The authors identified motives for consumers to generate eWOM on online ‘consumer-opinion platforms’. Their theoretical framework is based on an earlier theory by Balasubramanian and Mahajan (2001) and builds on it. Two additional

motivational drivers that Balasubramanian and Mahajan (2001) came up with are focus-related

utility and consumption utility. Focus-related value refers to the consumer’s belief that they add

(19)

12 previously mentioned antecedents: reputation, sense of belonging, and enjoyment of helping.

Consumption utility refers to consumers who share and obtain benefits through ‘direct

consumption of the contributions of other community constituents’ (Balasubramanian & Mahajan, 2001). Consumers give their opinion on the basis of consumption experience, which

others can then make use of. Besides the joy of helping others there is also the need to help

others if negative experience comes into play. Out of concern for others, reviewers may post

a review online to protect other members of the community (Hennig-Thurau et al., 2004). The

motivations of consumers to create UGC are generally based on “ego-defence and

social-functional” reasons. Ego-defence stands for self-protection, or the mitigation of uncertainty and

hazards. Social-functional reasons, in turn, are about helping consumers to reduce their sense

of self-doubt and creating a feeling of belonging to a specific society. The social function of

UGC creates a benign feeling and the ability to identify oneself with others. Furthermore, it

generates a sense of being someone (Daugherty et al., 2008).

Another writing motivation is the gathering of information. Consumers start consuming

after they read reviews, written by others, on a web-based opinion platform. Post-consumption,

consumers may be motivated to comment on those reviews. Consumers review to share their

experiences, but also to ask for solutions to specific problems or complaints. This way of

advice-seeking before making a purchase is “concerned with acquiring the skills necessary to

better understand, use, operate, modify, and/or repair a product” (Balasubramanian & Mahajan, 2001; Hennig-Thurau et al., 2004). As an extension, Hennig-Thurau et al. (2004)

added two new utilities to the three listed by Balasubramanian and Mahajan (2001). These are

the moderator-related utility and the homeostase utility. The moderator-related utility refers to

the mediating role that reviews and complaints may take up between the customer and the

company. This mediating role is about convenience and problem-solving support. A consumer

(20)

13 utility, in turn, is concerned with the desire to restore an equilibrium. According to the balance theory, people have a basic need to find balance in their life-time (Heider, 1946; Hennig-Thurau

et al., 2004). This could work on two sides. Firstly, if a consumer is left with an extremely

positive feeling of satisfaction after consumption, they may desire to share the joy they received

from using the product or service (Dichter, 1966). Similarly, after a bad experience with a

product or service, a consumer may write a review on a platform to reduce their feeling of

frustration and retain balance (Hennig-Thurau et al., 2004).

Hennig-Thurau et al. (2004) address two more reasons to write online: platform

assistance and economic incentives. Platform assistance occurs when someone believes that reviewing online will contribute to a company becoming more accommodating, or when it is

easier than writing or calling. Economic incentives refer to rewards and incentives offered for

writing online (Hennig-Thurau et al., 2004). This is the most common way in which companies

try to manipulate the reviews they receive online. They might ask the consumer to write a

positive review in return for a reward (ACM, 2017).

In summary, Hennig-Thurau et al. (2004) list a total of eight main motivations for

writing on opinion platforms. These are: (1) positive self-enhancement, (2) negative feelings,

(3) social aspects, (4) helping the company, (5) concern for others, (6) advice seeking, (7) platform assistance, and (8) economic incentives.

2.3 Sentiment in reviews

Besides the different motivations for why reviews are written, reviews may also differ

in sentiment (the degree to which a review is positive or negative). Specific motivations may

lead consumers to write reviews with specific intentions. For example, the motivations of

positive self-enhancement and helping the company would lead consumers to write reviews with a more positive sentiment than would the motivation of negative feelings. Apart from

(21)

14

2.3.1 Gender and sentiment

According to Wang, Burke and Kraut (2013), gender plays a big role in the way in which

content is generated on Facebook. The authors looked at sets of keywords related to different

topics, such as family, food or the weather, and attempted to find out if there was a difference

between genders in the frequency with which they talked about these topics. A few of the topics

they investigated can be related to reviewing restaurants: thankfulness, complaining, food,

negativity about people, and anticipation. The results showed that women on Facebook are

more likely to talk about thankfulness and anticipation compared to men. The popularity of

topics such as complaining and food is roughly equal between men and women. As for the topic ‘negativity about people’, it is clear that men have less difficulty starting a discussion about this topic than women (Wang, Burke, & Kraut, 2013).

Kim, Lehto and Morrison (2007) investigated what the differences are between the

genders in terms of their internet search behaviour, attitude and content preference on online

travel websites. One of the items the authors researched is restaurants. The results show that

women are more likely to search for restaurants and read information about restaurants. They

also show that the content women generate online is much more positive, a finding backed by

Kim et al. (2006). When women read information about travel and nightlife online, they read

most often about entertainment, restaurants and local information. Similar findings are reported

in a paper by Gretzel and Yoo (2008). They mention that “[g]ender differences were found for

perceived impacts of reviews, with females reaping greater benefits from using reviews, especially in terms of enjoyment and idea generation.”

2.3.2 Sentiment and expertise

Consumers with a higher level of expertise gather different information from those with

less expertise. Put concretely, an individual with more knowledge will tend to ask for more

(22)

15 Experts consider the characteristics of a product or service as informative, while novices its

stated benefits and drawbacks as more informative (Park & Kim, 2009; Walker, Celsi, & Olson,

1987). An example from Park and Kim (2009): “experts make judgments about food items on

the basis of technical attributes (e.g. nutritional information), but novices tend to use benefit information about the items (e.g. good for you).” This is one reason why reviews written by experts are more critical. Experts tend to address more small details than an average individual

or novice. (Vermeulen & Seegers, 2009). According to Lee, Law and Murphy (2011), traveling

expertise is also positively correlated with the helpfulness of the review. This implies that the

more expertise a reviewer has, the more useful the review will be for the readers. In general the

reviews written by experts are better substantiated. The same study also analysed the probability

of someone remaining an active reviewer on TripAdvisor. The results indicates that 45% of

reviewers keep contributing after writing their first review. In other words, more than 50%

never posted a review again after the first. Of those who did post a second review, 69%

remained active afterwards; and of those posted a third, 75% remained active. In sum, travel

experience leads to greater activity on TripAdvisor, while writing experience is important for

maintaining activity.

2.4 Differences between social media and online review platforms

In this subsection, a distinction will be made between social media and review

platforms, and it will be made clear what the differences are between these two platform types.

Social media platforms

There are many review platforms available for restaurants. The biggest of these is Yelp.

According to Yelp (2017), of the consumers who search online for local businesses, 74% turn

to review sites. Review platforms, such as Yelp, TripAdvisor and OpenTable, are in effect

online communities which gather to talk about a specific topic (in this case restaurants). The

(23)

16 consumers to make reservations. Yelp has even got a reviewers’ network were people can share

their restaurant experiences (Yelp, 2017b). The aim of online review platforms is essentially to

allow consumers to share their experiences with a product or service with other potential

consumers.

Lee, Law and Murphy (2011) examined online opinion leaders, or experts, on

TripAdvisor. Experts are people who possess experience with the product or service under

review. The authors found that reviews posted by experts are generally considered to be more

helpful by users lacking experience. Most online review platforms allow users to see which

reviewers are the most experienced. TripAdvisor, for example, lets users know how much

experience a reviewer has with traveling and writing reviews. The more a reviewer has travelled

and/or written reviews before, the more expertise they are considered to have. Their experience

makes them an expert and may also act as a stimulus for the rest of the travel community to

remain or become active travellers and reviewers. Helpful reviewers post often on TripAdvisor

because they want to help others in the community or because they receive appreciation from

the other consumers. Unlike most dedicated review platforms, Facebook does not display

information about the reviewer other than the review itself, preventing users from knowing who

the experts are. Moreover, online review platforms have guidelines and rules that determine

what reviews look like, which helps make the reviews more valuable. Facebook does not have

such review standards. Appendix 2 shows some examples of what a review form might look

like on an online restaurant review platform (TripAdvisor, 2017; Yelp, 2017a).

Review websites are a very important medium for restaurants. Earlier research found

that one extra star on a Yelp rating could lead to a 5-9 per cent increase in revenue for the

restaurant in question (Luca, 2016). Half a star improvement will improve the that the restaurant

(24)

17 Social media

There are multiple types of social media on which people generate content which present

their opinions. Sites like Twitter, Facebook and YouTube

make it possible to share such information. Facebook has

at the time of writing 1.94 billion active users. With that,

Facebook is the largest player regards in the social media

platform market, as shown in figure 1: ‘social media use

worldwide’ (Statista, 2017). Since 2015, Facebook has allowed users to review restaurants and give them star-based ratings (Mashable, 2015).

YouTube, Instagram, and Twitter, the other prominent social media platforms, allows only for

users to ‘like’ things or talk about subjects. Exclusively among the social media platforms,

Facebook makes it possible to post real restaurant reviews with star ratings. In comparison with

other social media websites, Facebook’s review features look more like those on online review platforms.

A previous study by (Kietzmann, Hermkens, McCarthy, & Silvestre, 2011) aimed to

catalogue the different functionalities of social media. They state that Facebook allows users to

identify themselves by means of a personal profile, show relationships, share and receive

information, start conversations with groups or individuals, build a reputation among people

they knew, and share content. Gonzales and Hancock (2011) investigated what happens with users’ self-esteem when they edit their personal information on Facebook. Their findings show that the personal profile acts as a ‘unique awareness-enhance stimuli’. Put differently, the

personal profile functions like a ‘mirror’. Because the information shared in the profile is based

on personal ‘stories’, the profile generally paints a more positive picture than reality. People are more likely to share positive information about themselves than negative.

(25)

18 The major difference between social media and online review platforms is that online

review platforms behave more like communities concentrated around a single topic, for

example travel or restaurants. Their users are interested in information and reviews about this

specific topic. Review platforms also tend to display the expert status of the reviewer to inform

other users of their reviewing experience (Lee et al., 2011). Moreover, on review platforms

UGC is shared most often with strangers, while UGC on Facebook is mostly shared with

friends. Hence, Facebook can be characterised as a platform for sharing information to close

relations like family and (potential) friends (Gonzales & Hancock, 2011; Lee et al., 2011). In

general, users have met most of the contacts they have befriended on Facebook in an offline

context. Facebook users share more personal information and their profiles function more as

self-presentations than those on review websites (Ellison, Steinfield, & Lampe, 2007;

Kietzmann et al., 2011; J. Kim & Lee, 2011; Lee-Won, Shim, Joo, & Park, 2014). Furthermore,

online review platforms employ guidelines to help users write quality reviews, while Facebook

does not.

2.5 Hypotheses and conceptual model

The hypotheses formulated in this section are derived from the literature reviewed

above. They match each of the writing motivations of online reviewers with either social media

or dedicated review platforms. This is done with the presumption that each type of platform has

certain characteristics that make them more conducive to certain writing motivations and less

conducive to others. Furthermore, within each type of platform, it is expected that each writing

motivation is coupled with either a positive or negative intention. As such, the hypotheses make

predictions about the relations between, on the one hand, the most suitable writing motivations

for each platform, and on the other hand the review intention (positive or negative) they

(26)

19

2.5.1 Writing motivations

It is expected that Facebook reviews will in general be more positive (compared to

review platforms) because Facebook users want to share their experiences with restaurants with

friends and others in their own personal circles (Ellison et al., 2007). Because on Facebook

most content is shared with familiar people, and because people wish to build a reputation

within that group, ego-defence and positive self-enhancement are expected to be important

writing motivations on Facebook (Cheung & Lee, 2012; Gangadharbatla, 2008).

Therefore the following hypotheses are formulated:

H1a: There is a positive relation between writing a review with positive intention and the

motivation positive self-enhancement on Facebook.

H1b: There is a positive relation between writing a review with negative intention and the

motivation ‘negative feelings’ on Facebook.

Because consumers who write reviews on review platforms are more familiar with the

communities on those websites and wish to be part of it, they share their experience and joy

with this community and feel the need to belong (social aspects to the community)

(Hennig-Thurau et al., 2004; Lakhani & Von Hippel, 2003). Therefore the following hypotheses are

formulated:

H1c: There is a positive relation between writing a review with positive intention and the

motivation ‘social aspects’ (Balasubramanian & Mahajan, 2001; Hennig-Thurau et al., 2004) on online review platforms.

H1d: There is a positive relation between writing a review with positive intention and the

motivation ‘helping the company’ on online review platform.

Besides sharing in the joy of others, consumers may also feel the need to help others.

(27)

20 part of the community, or consumers who make use of the reviews written by the community.

Therefore the following hypothesis is formulated:

H1e: There is a positive relation between writing a review with negative intention and the

motivation ‘concern for others’ on online review platform.

Because it is becoming increasingly easy to access information supplied by the

community of restaurant-goers, consumers are increasingly making use of online review

platforms to perform a mediating role. In doing so they receive problem-solving support from

other consumers or the restaurant (Balasubramanian & Mahajan, 2001; Hennig-Thurau et al.,

2004). Therefore the following hypothesis is formulated:

H1f: There is a positive relation between writing with positive intention and the motivation

‘advice seeking’, especially on online review platforms.

So, in conclusion: it is expected that negative feelings and positive self-enhancement are

the main motivations for writing a review on Facebook, while social aspects, concern for

others, helping the company, and advice seeking are the main writing motivations on online review platforms. This distinction is based on the characteristics of the platforms versus those

of social media. The motivations of social aspects, concern for others, helping the company,

and advice seeking are all related to helping others with the same interests, which in turn

corresponds most closely to the function of review platforms. In contrast, users on Facebook

operate more along the lines of social-functional reasoning, which corresponds to the

motivations of negative feelings and positive self-enhancement.

As for the last two motivations — economic incentives and platform assistance

(Hennig-Thurau et al., 2004) — the expectation is that there is no difference between the two types of

(28)

21 comes into play when the review platform is more effective than writing or calling. Since both

platform types work the same, no difference is expected here. Economic incentives are a factor

when reviewers are rewarded, but since no evidence was found to suggest one of the platform

types offers more attractive rewards than the others, no difference is expected here either.

H1g: ‘Economic incentives’ is not a motivation that leads to a positive or negative intention on

either type of platform.

H1h: ‘Platform assistance’ is not a motivation that leads to a positive or negative intention on

either type of platform.

Control variables

For controlling the review intention (positive or negative), control variables are used.

The frequency with which users go out to dinner (the more one eats out, the higher the expertise)

is used as a control variable (Park & Kim, 2009; Vermeulen & Seegers, 2009), because this

factor could influence the intention of the review. The same goes for the user’s experience in

writing reviews and their gender (Gretzel & Yoo, 2008; D. Kim et al., 2007).

2.5.2 Sentiment

It is expected that there will be differences in sentiment between women and men,

because women tend to behave more positively online than men (Gretzel & Yoo, 2008; D. Kim

et al., 2007). Therefore the following hypothesis is formulated:

H2: Women write reviews with a more positive sentiment compared to men.

Because reviewers supply both a star rating and a written review of their experience, it

is predicted that star rating and review sentiment are related to each other. Furthermore, because

users tend to share more personal information on social media, it is expected that Facebook will

(29)

22 Hancock, 2011). Because the communities on online review platforms are based around

restaurant-going, their users will be more nuanced and more critical in their reviews (Ellison et

al., 2007; Kietzmann et al., 2011; J. Kim & Lee, 2011; Lee-Won et al., 2014).

H3: There is a positive relation between stars and sentiment for both types of platforms.

H4: On Facebook the sentiment score of the reviews will be higher compared to online review

platforms.

2.5.3 Conceptual model

Experience (did the user ever write a review), frequency (expertise in eating out) and

gender are the control variables of the variable review intention.

Review intention may breed a specific sentiment and a star rating. However, no such causal

relation was in the present study (see chapter 3; methods), and therefore this relation is ignored.

However, a relation is expected to be found between star rating and sentiment. Figure 2 shows

the generated conceptual model:

(30)

23

3. Methods

This study is founded upon two different datasets which are used to answer the research

question: To what extent is there a difference between online review platforms and social media

(Facebook) in terms of writing motivations and sentiment of online reviews? The first dataset was obtained through a questionnaire among Dutch citizens which inquires after their main

writing motivations for online restaurant reviews. The second dataset is a collection of online

restaurant reviews that consumers posted on Facebook and Yelp (an online review platform for

reviewing restaurants). All the reviews collected pertain to the same six restaurants. An analysis

of the two datasets will make it possible to find out: (1) what the main reasons are ‘why’ people

write online reviews and (2) if there are differences in sentiment score between the platforms.

3.1 Dataset 1: Questionnaire

This section covers the questionnaire that constitutes he first dataset. First the collection

of the data will be discussed, followed by the assimilation and analysis of the data.

3.1.1 The data collection

The first dataset, the online questionnaire, will aid in answering the question “what motivates people to post a review on Facebook or on a regular online platform?” Respondents were asked to give their opinion on eight possible motivations for why they might write an

online review on Facebook or TripAdvisor. The present study is only focused on Facebook,

because it is the largest social media platform worldwide (figure 1). TripAdvisor was

nevertheless included because the website utilises a clear review form with clear filling-in

procedures. This helped to highlight the differences between Facebook and online review

platforms.

The data was collected through an online questionnaire with 50 statements (appendix

(31)

24 Thornhill, 2011). The 50 obtained statements were formulated with the help of existing

literature to achieve a higher Cronbach’s alpha. The statements were selected from previously

tested frameworks which were used as a foundation for the questionnaire. The surveys were

obtained through personal contact with their authors. Three questionnaires were thus obtained.

Of those three questionnaires, two — the one by Gangadharbatla (2008) and the one by de Hoog

(2011) — were used. The final questionnaire was then created using these two surveys in

combination with the scales developed by Hennig-Thurau (2004). All three researchers

examined consumers’ motivations for posting reviews online. Collectively, 8 main motivations

can be derived from their findings: “platform assistance, venting negative feelings, concern of

other consumers, positive self-enhancement, social motives, economic incentives, helping the company, and advice seeking” (Bronner & de Hoog, 2011; Gangadharbatla, 2008; Hennig-Thurau et al., 2004).

The quantitative approach makes data collection easier and allows for a better

understanding of which variables influence the dependent variable of review intention

(Saunders et al., 2011). To make a distinction between the common writing motivations on

either type of platform, the respondents were presented with statements on what they consider

to be important motivational factors for writing online reviews. The statements were prefaced

with a clear request to the respondents to think about why they write reviews on Facebook.

Additionally, a picture was shown of a review writing form. The same was repeated for

TripAdvisor reviews.

A 5-point Likert scale was used for the 50 motivational statements. The first

questionnaire was in English, as the literature used to create it was written in English too.

However, to make the questionnaire more suited to Dutch speakers, it was also translated to

(32)

25 English and Dutch speaker to make sure the statements were correctly translated (Dörnyei &

Taguchi, 2009).

The variables tested for in the questionnaire are listed and described in table 1:

‘collected variables of the questionnaire’.

The survey was distributed via qualtrics.com. Respondents were gathered in several

different ways. Firstly, social media was used to ask the author’s friends, family and

acquaintances to participate via private message. In addition, the online Facebook page ‘collecting respondents’ was used. This page brings together students who wish to share their surveys with each other. Secondly, respondents were recruited on the street. This was used as

an additional method to improve the likelihood that the data would be correct and complete.

Respondents were offered sweets in return for participation as an incentive. All surveys filled

Variable Description Platform

(independent variable):

A distinction was made between TripAdvisor and Facebook (comparison between online review platforms and social media).

Motivation (independent variables):

These are possible motivations for why a user might write online reviews:

1. Advise seeking: Someone posts reviews in the hope of obtaining advice from others.

2. Positive self-enhancement: Someone posts reviews to share their personal joy as a result of the reviewed experience.

3. Negative feelings: Someone posts reviews to share their negative experience with the reviewed product/service.

4. Concern for others: Someone posts reviews to help others.

5. Social aspects: Someone posts reviews because they wish to belong to a community

6. Economic incentives: Someone posts reviews in order to obtain a reward. 7. Helping the company: Someone posts reviews to help the company

(restaurant) improve.

8. Platform assistance: Someone uses the platform to assist themselves, to receive a quicker reply from the company.

Gender

(control variable):

Male and female

Frequency (control variable):

How many times do you go out for dinner? 1. Once every week

2. Once every month 3. Once every three months 4. Every half a year 5. Every year

Experience (control variable):

Whether someone has experience with writing reviews online: yes or no

Intention

(dependent variable):

Intention of someone’s reviews: If he/she writes online review, is this usually with a positive or with a negative intention?

(33)

26 in this way were filled in for 100%. Thirdly, respondents were gathered by inducing a snowball

effect (Biernacki & Waldorf, 1981). Friends of the author were asked to share the survey within

the companies they worked for, and also to share the survey via LinkedIn the Facebook page

of Hogeschool Utrecht, which has more than 25,000 followers. A total of 258 respondents were

recruited this way. All data was transferred to SPSS for analysis.1

3.1.2 Pre-test

A pre-test was undertaken to evaluate whether the statements on the questionnaire were

interpreted correctly. According to Saunders et al. (2011) a sample of 30 is sufficient for a

pre-test. After the first 30 respondents had completed the survey the pre-test was executed. First,

the drop-out rate was analysed. Drop-out data was retrieved from Qualtrics, which records the

point at which respondents stop filling in the questionnaire. The initial questionnaire presented

all the statements about Facebook, whereupon the respondent was requested to fill in the same

statements, but for TripAdvisor. This was experienced as boring, so the dropout rate was higher

than expected. To make the questionnaire less boring, an altered version of the questionnaire

first presented five or six statements about Facebook, then the same statements but for

TripAdvisor, then more statements about Facebook, and so on. This made it easier to fill in,

and less boring. Also, this alternating way of asking questions encourages the respondent to

think about the differences between the platforms.

Secondly, the Cronbach’s Alpha of the pre-test was checked. In the pre-test all the

writing motivation statements scored higher than 0.7, which is sufficient (Field, 2009).

1_{For this research the software program IBM SPSS Statistics was used. This program is supported by the} University of Amsterdam. Hence a valid license for SPSS was obtained.

(34)

27

3.1.3 Assimilation of the data

At first the sample consisted of 258 respondents. The first step was removing all the

respondents who did not complete the entire survey. Next, people who completed the survey

within 120 seconds were also taken out of consideration, since it was thought unlikely that

anyone could properly complete the survey within two minutes. After these respondents were

removed, the remaining total was 217.

Recoding was also part of cleaning the data. This involved, for example, changing 1

(male) and 2 (female) into 0 (male and 1 (female), as well as recoding counter-indicative items.

Dummies were coded to create a clear distinction between positive and negative reviews.

Dummies were also created for experience, encoding whether respondents were either

experienced in writing reviews (coded as 1) or not (coded as 0). This was done because the

large majority of respondents had written either zero reviews or between one and five reviews

(80.2%). Due to the extreme differences between group sizes, the results would not have been

representative. Therefore experience in online reviewing was coded into ‘has experience’, and ‘does not have experience’.

Furthermore, age was not taken into account since the results were not normally

distributed in terms of age. The group of 18-30 years old made up almost 92% of respondents.

3.1.4 Reliability analysis

Each writing motivation was tested using two or three items. Reliability analysis was

carried out to establish whether those items measured the construct they were designed to test

for. Almost all constructs scored 0.7 or higher on the Cronbach’s alpha (between 0.692 and

0.924, see also table 6 and 7), except for the motivation platform assistance (on Facebook and

TripAdvisor). This construct had a Cronbach’s alpha of 0.548 for Facebook and 0.565 for

TripAdvisor. Some variables hovered around 0.7, but some existing literature agrees that a

(35)

28 assistance as a writing motivation was left out of the rest of the analysis. For the motivations positive self-enhancement and social aspects on Facebook, one of the three respective items had to be removed to achieve a higher reliability.

3.1.5 Method of analysis

Several analyses were carried out to learn more about the factors influencing review

intention. The first analysis looked at the descriptive statistics of all the variables to identify all

the means and outliers. Compared samples T-tests were also carried out, to identify possible

differences between the two platform types and writing motivation.

Subsequently, a correlation analysis was undertaken to establish correlation among

variables. Next, multicollinearity analysis was run to make sure there were no disruptive

collinearities. The Variance Inflation Factor (VIF), which is used to measure the degree of

multicollinearity, needs to be below the value of 10 to make sure independent and control

variables are not of critical influence (Field, 2009; O’brien, 2007). This is an essential

prerequisite for carrying out a logistic regression — if not met, it cannot be said with certainty

whether the dependent variable is explained by the model or merely by some extreme

correlating variables.

The logistic regression indicates whether the writing motivations and control variables

play a role in affecting review intention. The logistic regression approach fits best, because the

dependent variable is dichotomous (Field, 2009; Hosmer Jr, Lemeshow, & Sturdivant, 2013).

A distinction was made between reviews written with a positive intention and those written

with negative intention. Thus, a review was considered either negative or not negative, the same

applying to positive variables. The results of the analysis are presented in chapter 4, ‘Results of

(36)

29

3.2 Dataset 2: Review database

This section is about second dataset, that of the online reviews. First, the manner in

which this dataset was collected will be discussed. This is followed by an explanation of how

the data was assimilated and a description of the subsequent data analysis.

3.2.1 Data collection

The first step was the random selection of six restaurants on Yelp for which to collect

the reviews. The search was limited to just regular restaurants (so no star-awarded restaurants,

fast food restaurant, et cetera). The first restaurant on every 10th_{page of the Yelp search results}

was chosen. A restaurant was eligible for selection if its Facebook page had at least 50 reviews

on it. If this condition was not met, the first restaurant on the next page was selected. This

amounted to random selection, which is important in order to obtain a sample that is as

representative of the population as possible (Saunders et al., 2011).

New York was selected as the location of the selected restaurants. New York was

primarily chosen because it is home to a great many restaurants to choose, with a high diversity

to boot. Tripadvisor.com alone lists 12,500 restaurants in New York. Yelp.com lists even more

New York restaurants at 40,000. Another reason for choosing New York is that no research has

been done on New York restaurants before. Additionally, the choice of New York allowed for

the exclusion of certain influence factors that might arise as a result of cultural differences

between the United States and the Netherlands (as the questionnaire respondents were Dutch).

Cultural factors and accessibility to review platforms are known to be important drivers for

customers to post a review (Berthon, Pitt, & Campbell, 2008; Shao, 2009; Smith, Fischer, &

Yongjian, 2012). New York hosts many restaurants, many of which are more or less

look-a-likes of European restaurants, helping to close the cultural gap between New York and the

Netherlands (Browne & Browne, 2001). Furthermore, like the Netherlands, New York is part

(37)

30 internet literate. New Yorkers are likely to have a similar degree of access to review platforms

as Dutch citizens.

Reviews were collected from Yelp because it was compatible with the scraping tool

Webharvy. Webharvy aids the researcher in scraping of data off the internet, and of all the

platforms could only be used with Yelp reviews. The choice of Yelp is reinforced by the fact

that Yelp has more restaurant reviews compared to TripAdvisor. Facebook reviews were

collected manually.

The collected set of reviews, when analysed, grants insight into the sentiment polarity

of restaurant reviews on Facebook compared with those on Yelp. Semantria was used to

calculate the polarity of the reviews — how positive or negative reviews were. For this step,

quantitative content analysis of reviews posted online was carried out. Semantria was used to

extract sentiment from the reviews. When consumers read a review online, they can make a

distinction between positive and negative reviews. They can also judge whether one negative

review is more negative than another negative review. The Semantria sentiment analysis makes

the same judgment and scores the sentiment of the review between -1 to +1. Then it combines

this score with the sentiment of the individual sentences. The author refrained from doing the

scoring manually because unlike the author, the Semantria system is not affected by external

factors such as weather or mood. As such, the sentiment score given by Semantria is more

objective and consistent (Lexalytics, 2017). The fact that Semantria works better with English

reviews than with Dutch ones was another reason for analysing New York reviews and not

reviews from the Netherlands. The collected variables are in table 2: ‘variables of the review

The differences between reviews on Facebook and online review platforms : a link between online writing motivations and sentiment