The influence of e-WOM on hotel room booking intentions : An investigation on the effect of valence and volume & the moderating effect of individual level uncertainty avoidance

(1)

The influence of e-WOM on hotel room Booking Intentions:

An investigation on the effect of Valence and Volume & the moderating effect of

individual level Uncertainty Avoidance

Kirsten Kuijt

11142383

University of Amsterdam Faculty of Economics and Business

MSc. In Business Administration – Marketing track Supervisor: Dr. Mattison Thompson

Final version 23rd of June, 2017

(2)

| 2

Statement of originality

This document is written by Kirsten Kuijt 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 supervision of completion of the work, not for the contents.

(3)

The influence of e-WOM on hotel room booking intentions

| 3

Abstract

In the tourism and hospitality industry word-of-mouth, especially in the form of online reviews, is of great significance and influences consumers enormously. This research aimed to examine the influence of the valence and volume of online reviews and the interaction effects between the two on hotel room booking intentions. Furthermore, the effect of individual culture was taken into account. It was investigated if the level of uncertainty avoidance of a person moderates the influence of valence and volume on hotel room booking intentions. The experiment conducted was a 2x2 between-subjects factorial design.

Participants were exposed to one of the four conditions, which included few or many reviews and neutral or negative reviews. Results showed that review valence has a direct effect on hotel room booking intentions. Neutral reviews lead to significantly higher hotel room booking intentions than negative reviews. Review volume does not have a significant direct effect on hotel room booking intentions and no interaction effect was found between valence and volume. Also uncertainty avoidance does not moderate the effect of valence and volume on hotel room booking intentions.

Key words: e-WOM, online reviews, Valence, Volume, Uncertainty Avoidance, hotel room Booking Intentions

(4)

| 4

List of tables and figures

Tables

Table 1: 2x2 between subjects factorial design ... 20

Table 2: Demographics ... 25

Table 3: Mean scores of valence ... 26

Table 4: Mean scores of volume ... 27

Table 5: Means, Standard Deviations, Correlations and Reliabilities ... 28

Table 6: Factorial ANCOVA ... 30

Table 7: Moderation on the effect of valence on booking intentions ... 32

Table 8: Moderation on the effect of valence on booking intentions (2) ... 33

Table 9: Moderation on the effect of volume on booking intentions ... 33

Table 10: Moderation on the effect of volume on booking intentions (2) ... 33

Table 11: Summary of results ... 34

Table 12: ANOVA to test independence of treatment variable and covariate ... 56

Table 13: ANOVA to test independence of treatment variable and covariate (2) ... 56

Table 16: ANCOVA to test homogeneity of regression slopes ... 58

Figures Figure 1: Conceptual model ... 18

Figure 2: Information of experiment ... 20

Figure 3: Example of condition ... 21

Figure 4: Plot factorial ANCOVA ... 31

(7)

Introduction | 7

1. Introduction

Word-of-mouth, e-word-of-mouth or online reviews are terms that will sound familiar to many people, and many people read reviews before buying a product or service. The importance of word-of-mouth (WOM) has been recognized by many researchers (e.g.

Cheung, Anitsal, & Anitsal, 2007; Chevalier & Mayzlin, 2006; Day, 1971; Hennig-Thurau & Walsh, 2004; Mauri & Minazzi, 2013; Schumann et al., 2010). Originally, WOM was about person-to-person communication but nowadays people can reach a broader audience as a result of the emergence of the internet. WOM online is called electronic word-of-mouth (e-WOM). Especially in the tourism and hospitality industry e-WOM – in the form of online reviews – is of great significance. This is because services in the hospitality and tourism industry are considered experience goods and the quality of these goods is often not known before consumption (Ye, Law, Gu, & Chen, 2011). Consumers need to rely on experiences of others to obtain some information about the quality of the services they are interested in. It has been found that many people check the internet before booking a hotel room (Prabu, 2014). The receiver of WOM, in this case the reader of an online review, is affected by the information that the online review contains and may change his or her behaviour (Hennig-Thurau & Walsh 2004).

Several characteristics of online reviews can have an influence on hotel room booking intentions. Among other things, valence and volume of online reviews can have an effect and this is studied in this research. Some previous studies found that the valence of reviews

influences hotel room booking intentions (Ladhari & Michaud, 2015; Mauri & Minazzi, 2013; Ye et al., 2011). Specifically, negative reviews decrease hotel room booking intentions and positive reviews increase hotel room booking intentions. This research looks into the effect of negative and neutral reviews. Besides valence, volume can also be of importance for review readers. It has been found for example that volume of reviews has a positive effect on movie sales(Duan, Gu & Whinston, 2008). An interaction effect can also be expected. For example, Tsao, Hsieh, Shih and Lin (2015) found that positive reviews become more persuasive when the number of reviews increased. Negative reviews, in turn, led to more negative hotel room booking intentions. However, not much research has been executed in the hospitality industry concerning volume of online reviews.

There might also be factors that moderate the effect of online reviews on hotel room booking intentions. Not every person will respond the same way to online reviews. People differ in their cultural values, differences in cultural values arise between countries but also

(8)

Introduction | 8 within a country (Yoo & Donthu, 2002). Cultural values might have an effect on how online reviews are perceived and how consumers adapt their behaviour because of it (Schumann et al., 2010). People can differ on their level of uncertainty avoidance and it is expected that the level of uncertainty avoidance of a person is a cultural factor that has an influence on the relationship between online reviews and hotel room booking intentions. People high in uncertainty avoidance are described by Mooij and Hofstede (2011) as people who feel uncomfortable in new, unknown or surprising situations.

A lot of research has been done on e-WOM in the tourism and hospitality industry and the effect on hotel room booking intentions. However, most research looks into positive and negative reviews, but reviews can also be neutral. To the best of my knowledge, no study looked at neutral reviews and especially not in combination with the volume of online reviews. It might of interest to know what effect neutral reviews can have on booking

intentions and what happens if there is a high volume of neutral reviews. Also, the moderating effect of individual cultural differences (uncertainty avoidance) on the relationship between valence and volume on hotel room booking intentions has not been researched to the best of my knowledge. This research will try to fill this gap with the following research question:

What is the effect of online review valence and volume on hotel room booking intentions and what is the moderating effect of individual level uncertainty avoidance?

This research is structured as follows. First, a literature review will explain the main concepts and discusses existing literature concerning the topic of this research. The

conceptual framework that follows explains the relationship between the variables, covers the hypothesis and formulation of these and a conceptual model is given. The methodology explains the research design, measurement of variables and provides information about the survey. This is followed by the results that explain the sample, statistical procedure and the outcomes of the research. Last, the discussion reports the main findings, implications, limitations and provides suggestion for future research.

(9)

Literature review | 9

2. Literature review

This chapter discusses existing literature concerning the main concepts of this research. These are word-of-mouth, e-word-of-mouth, online reviews in the tourism and hospitality industry, the effect of online reviews, cultural factors, national versus individual culture and uncertainty avoidance. This chapter also describes the gap found in the literature and the research

question of this research. 2.1 Word-of-mouth (WOM)

Word-of-mouth or its abbreviation WOM is an age old subject and many attempts have been undertaken to define word-of-mouth communication. It was defined by Arndt (1967) as: “Oral, person-to-person communication between a receiver and a communicator whom the receiver perceives as non-commercial, regarding a brand, a product or a service” (Arndt, 1967 as cited in Buttle, 1998. p242). Compared to advertising, WOM occurs in real time and real life, it contains the exchange of verbal actions between a source and recipient as a response to particular circumstances (Stern, 1994). These verbal actions are personally motivated,

spontaneous, short-lived, informal (not commercial) and WOM communications disappear as soon as they are expressed (Stern, 1994). Simplified, WOM is a concept where a person shares his or her view with another person. It is agreed upon in the literature that WOM is very powerful and has a much greater influence on people than other forms of marketing communication (Day, 1971). WOM communications have an important role in shaping consumers’ attitudes and behaviours (Cheung et al., 2007).

2.2 Electronic word-of-mouth (e-WOM)

Nowadays, WOM extends to the internet, which is called electronic word-of-mouth or e-WOM. It differs from traditional WOM in the sense that it takes place in a more complex, online environment and not face-to-face and in private (King, Racherla, & Bush, 2014). As opposed to traditional WOM, e-WOM is not short lived and can reach an unlimited number of receivers who the sender may or may not know (Godes & Mayzlin, 2004).

According to Bickart and Schindler (2001) e-WOM has greater credibility, relevance and empathy than content created by marketers. Opinions on for example internet forums are seen as trustworthy since the writers of the opinions are consumers who do not need to manipulate the reader and persuade them in buying a certain product or service. Although participants on internet forums may be very different from the reader, they are equal

(10)

Literature review | 10 the information very relevant (Bickart & Schindler, 2001).

There are several types of e-WOM (websites, e-mail, blogs, online reviews,

newsgroups, chat rooms, etc.) but the focus of this research will be on online reviews.Online user reviews have become a significant source of information to consumers, substituting and complementing other forms of business-to-consumer and offline WOM communication (Chevalier & Mayzlin, 2006).

2.3 Online reviews in the tourism and hospitality industry

In the tourism and hospitality industry, online reviews are widely used by consumers in order to obtain information, and plan their travels (Litvin, Goldsmith, & Pan, 2008). A well-known example of a website containing online travel reviews is Tripadvisor.com. People can read how others reviewed destinations, restaurants, hotels, flights and trips. TripAdvisor describes itself as the world’s largest travel site with over 435 million sincere traveller reviews that help you make the right choice when looking for a hotel, restaurant or attraction (TripAdvisor, 2017). Reviews written by other travellers are perceived by readers to be up to date,

enjoyable, and more reliable than information that travel companies provide (Gretzel & Yoo, 2008). Already in 2007 one-third of travel purchasers consulted a type of e-WOM before they purchased online (Compete, 2007 in Ye et al., 2011) and according to a research by

PhoCusWright, 77% of TripAdvisor users always read online reviews before booking a hotel room (Prabu, 2014).

Online reviews are probably so popular for tourism and hospitality related services because most of these services are considered experience goods (Ye et al., 2011). The influence of online reviews is substantial for these so-called experience goods because the quality of the good is often not known before consumption, and consumers have to rely on WOM and online reviews to get some information about the quality (Ye et al., 2011). Booking a hotel room can thus be seen as a risk and reading online reviews can reduce this risk. For example, 83% of TripAdvisor users mention that reviews make them more confident in their decisions (Prabu, 2014). Furthermore, Kim, Mattila, and Baloglu (2011) found three reasons why consumers are reading online hotel reviews. These are convenience & quality, social reassurance and risk reduction. Mauri and Minazzi (2013) and Ye et al. (2011) emphasized that travel reviews are an important source of information which influences the customer decision-making process and booking intentions.

(11)

Literature review | 11 2.4 The effect of online reviews

It is clear that online reviews are an important source of information for travellers, but what effect do online reviews actually have on the behaviour of consumers? In general, according to Hennig-Thurau and Walsh (2004) it can be expected that behaviour changes after reading online reviews. Consumers actively search for buying-related information on opinion

platforms, due to the content read this will change their behaviour (Hennig-Thurau & Walsh, 2004). Several characteristics of online reviews can have an influence on behaviour, among other things valence and volume can have an effect.

2.4.1 Valence of online reviews

One of the characteristics of online reviews that can have an influence on hotel room booking intentions is valence. Valence refers to the message (in this case online review) being positive or negative (Buttle, 1998). Many researchers looked at the effect of valence of online reviews on sales, awareness, attitudes and booking intentions. Vermeulen and Seegers (2008) found that both positive and negative reviews increase consumer awareness of hotels, whereas positive reviews also improve attitudes towards hotels (Vermeulen & Seegers, 2008). Liu (2006) reports that positive reviews increase expected quality and negative reviews decrease expected quality. Ladhari and Michaud (2015); Mauri and Minazzi (2013); Sparks and

Browning (2011) and Ye et al. (2011) found that booking intentions in the hospitality industry and online sales of hotel rooms are influenced by the valence of online reviews. Positive comments increase purchase intentions and negative reviews decrease purchase intentions of hotel rooms. This is also in accordance with the research of Park and Lee (2009) who found that the effect of online reviews is greater for negative than for positive online reviews, thus negative information can have a stronger negative effect on purchase decisions. Furthermore, they found that this effect is also greater for experience goods than for search goods (those goods of which complete information about the goods can be acquired before the purchase). Uncertainty about experience goods (that comes from the lack of knowledge before

consumption of the good) will be increased by negative online reviews (Park & Lee, 2009). It is clear that many researchers studied the effect of positive and negative reviews, but not many researchers looked at neutral or moderate reviews. Neutral or moderate reviews are reviews that are not extremely positive or negative. Mudambi and Schuff (2010) found that for experience goods moderate reviews are more helpful than extreme reviews since moderate reviews enhance the credibility of the review. Also, consumers are more open to moderate reviews since these are seen as more objective (Mudambi & Schuff, 2010).

(12)

Literature review | 12

2.4.2 Volume of online reviews

Besides online review valence, another characteristic of online reviews that can have an effect on behaviour is the volume of online reviews. Online review volume is the total amount of online review interaction (Liu, 2006) or, in other words, it is the total number of online reviews that is available. Several studies researched the volume of online reviews, but this is not researched often in the tourism and hospitality industry. For example, Duan et al. (2008) studied movie sales and found that sales are significantly influenced by the volume of online reviews. A higher number of reviews resulted in higher movie sales. Vermeulen and Seegers (2008) remark that one negative online review does not cause much problems, but what happens when there are multiple negative online reviews? A study that did look into volume in the tourism and hospitality industry is that of Tsao et al. (2015). They researched volume as a moderator between valence and hotel room booking intentions and found that positive reviews became more persuasive when the number of reviews increased. Negative reviews, in turn, led to more negative hotel room booking intentions when the number of reviews

increased.

2.5 Cultural factors and online reviews

Besides the abovementioned effects that characteristics of online reviews can have, cultural factors can also have an influence on how online reviews are perceived. According to Lo (2012) cultural values of a consumer play a role in the communication process of WOM. The action of the receiver can be guided by the value of the culture from which he or she belongs (Lo, 2012). Schumann et al. (2010) state that cultural values moderate the cognitive

processing of received WOM and, thus, the relevance that customers attribute to received WOM.

But what does culture mean? Hofstede defined it as: “the collective programming of the mind that distinguishes the members of one group or category of people from another” (Hofstede, 2001. p9). Hofstede studied the impact of different cultural values on management and his work is the basis for much cross-cultural research (Litvin, Crotts, & Hefner, 2004). He assessed 66 countries and developed cultural index scores and rankings for five ‘value’

dimensions on which countries differed: power distance, uncertainty avoidance,

individualism/collectivism, masculinity/femininity and long term orientation (Hofstede, 2001). A description of all dimensions can be found in Appendix 8.1.

(13)

Literature review | 13 2.6 National versus individual culture

Most previous research focuses on cross-cultural or national cultural differences, in which all members of a country are treated as if they are similar in culture (Yoo & Donthu, 2002). But national boundaries do not always characterize the cultural values of people, especially when the country has a heterogeneous population with different cultural backgrounds (Yoo & Donthu, 2002). For example, a person may live in a culture of high uncertainty avoidance, but can be low in uncertainty avoidance and might not be influenced by the high uncertainty culture he or she lives in (Donthu & Yoo, 1998). Culture can also be measured on an individual level. This is important since it may not work to only look at national culture and use that to target individual consumers (Yoo, Donthu, & Lenartowicz, 2011), diversity within a country can be greater than the diversity between countries (Donthu & Yoo, 1998).

Targeting individual consumers on the basis of national culture becomes less meaningful in today’s diversity among members in a nation and worldwide communication channels (Yoo et al., 2011). The dimensions of Hofstede (2001) can also be used to measure individual cultural values since it was originally developed on the basis of individual managers (Donthu & Yoo, 1998).

2.7 Uncertainty avoidance and its moderating effect

The focus of this research is on uncertainty avoidance and this is described by Hofstede (1994, p5) as: “the extent to which a culture programs its members to feel either

uncomfortable or comfortable in unstructured situations. Unstructured situations are novel, unknown, surprising and different from usual.” Uncertainty can come from risk, which can be described as a high chance of failure of an event, and from ambiguity, which can be described as the unknown chance of occurrence of an event (Donthu &Yoo, 1998). Risk or ambiguity can create uncertainty with which consumers feel uncomfortable (Donthu & Yoo, 1998). Consumers that are high in uncertainty avoidance are cautious in choosing services, take time to evaluate and do not take decisions quickly (Donthu & Yoo, 1998). Uncertainty avoidance is the dimension that is most widely used in the literature on consumer behaviour (Frías, Rodriquez, Castaneda & Sabiote, 2011). This is because factors such as risk and trust are among the most important determinants of consumers’ purchasing behaviour (Cheung, Chan & Limayem., 2005).And as mentioned before, one of the reasons why consumer are reading online reviews is risk reduction (Kim et al., 2011).

An example of a study that looked into the effects of cultural factors on consumer behaviour is that of Schumann et al. (2010). These researchers studied cross-cultural

(14)

Literature review | 14 differences and the effect of received word-of-mouth referral in relational service exchange. They found that received positive WOM has a positive effect on the perceived service quality. Also received WOM has a stronger effect on the evaluation of customers that are high in uncertainty-avoidance than customers who are low in uncertainty avoidance (Schumann et al., 2010). They also studied the other four value dimensions but only uncertainty avoidance proved a significant moderator. It is expected that the other cultural value dimensions do not have a moderating effect on the relationship between e-WOM and the decision making

process of consumers, especially in the tourism and hospitality industry. The definitions of the cultural dimensions show no indication that it would moderate the relationship between e-WOM and booking a hotel room.

2.8 Research gap

A lot of research has been executed on e-WOM in the hotel industry and the effect it can have on hotel room booking intentions. However, to the best of my knowledge, no study looked at neutral reviews in the tourism industry and its effect on hotel room booking intentions. Furthermore, research concerning volume of online reviews in the hospitality industry is scarce and volume of online reviews is not researched in combination with neutral online reviews. Also, I believe the moderating effect of individual cultural differences on the relationship between online reviews and hotel room booking intentions hasn’t been researched. This research will try to fill this gap.

This research contributes to existing literature by researching neutral online reviews and volume of online reviews in the hospitality industry. Mauri and Minazzi (2013) indicate in their research that an interesting point of research would be how behaviours of people of different cultures are influenced by online reviews. Although most journal articles investigate cross-cultural differences (Yoo & Donthu, 2002) it would also be of interest to research individual level cultures (cultural differences within a country). This research will add to the study of Schumann et al. (2010) by researching the effect of negative online reviews and the moderating effect of uncertainty avoidance in the service industry.

This research will investigate the effects of online review valence and volume on hotel room booking intentions. Furthermore, uncertainty avoidance (cultural value) will be

researched at an individual level as a moderator. The following question is researched in order to add knowledge to the tourism en e-WOM literature: What is the effect of online review

valence and volume on hotel room booking intentions and what is the moderating effect of individual level uncertainty avoidance

(15)

3. Conceptual framework

This chapter explains the relationships between the variables of this research. The chapter contains the hypotheses, the reasoning and theories behind it and the conceptual model is included.

3.1 Relationships and hypotheses

Nowadays the presence of online reviews has become very common. Most hotels can be found on websites such as TripAdvisor and many people read online reviews before they book a hotel. In general, it can be expected that reading online reviews can change consumer behaviour (Hennig-Thurau & Walsh, 2004). As mentioned in the literature review, online reviews influence the awareness, opinions, attitudes and booking intentions of consumers.

3.1.1 Valence of online reviews

One of the characteristics of online reviews that can have an influence on hotel room booking intentions is the valence of an online review. Valence refers to the positivity or negativity of the message, in this case online review (Buttle, 1998). Several researchers found that positive reviews are more effective than negative reviews in enhancing hotel room booking intentions (Ladhari & Michaud, 2015; Mauri & Minazzi, 2013; Tsao et al., 2015; Ye et al., 2011). Negative reviews often result in decreased hotel booking intentions (Mauri & Minazzi, 2013). This can be explained with the negativity effect or negativity bias. The negativity bias is a finding in impression formation literature, namely that people place greater weight on negative information than on positive information (Skowronski & Carlston, 1989). Negative information is more attention grabbing and easier to perceive than positive information and can therefore have a stronger effect on purchase intentions (Park & Lee, 2009). Negative reviews also augment the uncertainty and fear of consumers which makes them less likely to purchase or book a product or service, especially for experience goods (Park & Lee, 2009). Hence, negative WOM has a stronger influence on hotel room booking intentions than positive WOM (Park & Lee, 2009) .

Besides positive or negative, online reviews can also be neutral, which means that they are not extremely positive or negative. The extremity effect, which is similar to the negativity bias and originates from impression formation literature as well, reports that people place greater weight on extreme information than on moderate information (Skowronski &

Carlston, 1989). As mentioned, Mudambi and Schuff (2010) found that credibility of reviews increases when there are some moderate reviews in the review set, it also increases objectivity or the reviews. Thus, neutral online reviews are expected to have a less strong effect on hotel

(16)

Conceptual framework | 16 room booking intentions than negative online reviews and are more credible and helpful. Therefore the following hypothesis suggests that neutral reviews lead to higher hotel room booking intentions than negative reviews do.

H1: Valence of online reviews has a direct effect on hotel room booking intentions. The effect is more positive for neutral reviews than for negative reviews.

3.1.2 Volume of online reviews

Another characteristic of online reviews that can have an influence on hotel room booking intentions is the volume of online reviews. As described in the literature review, volume means the total number of online reviews that is available. The effect that volume of online reviews has on hotel room booking intentions can be explained with the awareness effect. A high number of online reviews has an informative function and increases the awareness of consumers about products or services and in turn this influences purchase or booking

intentions (Chen, Wang, & Xie, 2011; Liu, 2006; Park, Lee & Han, 2007; Tsao et al., 2015). According to Tsao et al. (2015) volume influences purchase intentions since the number of online reviews shows if and how many people are purchasing a certain product or service. When people know that many people purchased a product or service this rationalizes their own purchase of that product or service (Park et al., 2007).

When consumers are aware of the existence of a product or service they might put it in their choice set (Duan et al., 2008). This accords with the consideration set framework, in which awareness is an important term (Roberts & Lattin, 1991). Consumers narrow down all possible choice options to the awareness set, which is the set of options that consumers recall (Vermeulen & Seegers, 2008). The awareness set is then narrowed down to the consideration set (Vermeulen & Seegers, 2008). So, the consideration set framework proposes that

consumers might consider hotels that they are aware of (Vermeulen & Seegers, 2008), hotels that they are not aware of will not be in the awareness set, and consequently, not in the consideration set. Vermeulen and Seegers (2008) found that consideration significantly increases after respondents were exposed to an online review. Hence, a high volume of online reviews leads to a higher awareness of a hotel, and higher awareness leads to consideration (higher intentions to book). An example of a study that is in line with this theory is that of Liu (2006) in the movie industry, this study found that film revenues can be significantly

(17)

Conceptual framework | 17

H2: Volume of online reviews has a direct effect on hotel room booking intentions. The effect is more positive for many reviews than for few reviews.

3.1.3 Interaction effect

Apart from direct effects of review valence and volume, an interaction effect can also be expected. As mentioned, it is expected that neutral online reviews result in higher hotel room booking intentions than negative online reviews and that the number of reviews increases awareness and consequently increases consideration. However, Tsao et al. (2015) found that repetitive exposure to negative reviews is damaging to hotel room booking intentions, which can be explained by the negativity effect that suggests that people place more weight on negative online reviews (Skowronski & Carlston, 1998). They further found that the influence of valence on hotel room booking intentions is strengthened with the number of reviews, which in their case meant that many positive reviews lead to higher hotel room booking intentions than few positive reviews and many negative reviews lead to lower hotel room booking intentions than few negative reviews (Tsao et al., 2015).

H3: There is an interaction effect between valence and volume of online reviews on hotel room booking intentions. The effect of valence on hotel room booking intentions is strengthened with a higher volume of reviews.

3.1.4 Uncertainty avoidance

As mentioned in the literature review, cultural values can also have an effect on how online reviews are perceived. Uncertainty avoidance is the value that most likely has an effect on hotel room booking intentions. It is hard to know about the quality of a hotel before consumption, and therefore booking a hotel room can be seen as risky. A way to find out about the quality of a hotel beforehand is by searching for online reviews of people who visited the hotel. Sparks and Browning (2011) argue that online reviews can play an important part in reducing uncertainty. People high in uncertainty avoidance might be more susceptible to online reviews than people who are low in uncertainty avoidance (Schumann et al., 2010). It is therefore expected that the effect of valence of online reviews on hotel room booking intentions is stronger for people high in uncertainty avoidance and therefore the hypothesis is:

H4a. The effect of valence of online reviews on booking intentions is moderated by individual level uncertainty avoidance. The effect of valence of online reviews on hotel room booking

(18)

Conceptual framework | 18

intentions is stronger for people high in uncertainty avoidance than for people low in uncertainty avoidance.

People high in uncertainty avoidance will probably search for more information beforehand to avoid uncertainty. This is in line with the research of Money, Gilly, and Graham (1998) who found that business people in Japan (very high uncertainty avoidance) look for more

information sources in choosing a service provider than people in the United States (lower in uncertainty avoidance) do. More available reviews allow people to read more and this will especially satisfy people high in uncertainty avoidance. Therefore, it is assumed that people high in uncertainty avoidance will be influenced more by the volume of online reviews and the following hypothesis is derived:

H4b. The effect of volume of online reviews on hotel room booking intentions is moderated by individual level uncertainty avoidance. The effect of volume of online reviews on hotel room booking intentions is stronger for people high in uncertainty avoidance than for people low in uncertainty avoidance.

3.2 Conceptual model

The research question is visualized in a conceptual model (Figure 1) and enhances the understanding of how variables are related in this research.

(19)

Methodology | 19

4. Methodology

To investigate the research question a 2x2 between-subject experiment was executed. This chapter explains the research design, the measurement of variables and how the data was collected.

4.1 Research design

A deductive quantitative approach was chosen to pursue the research purposes, which was executed through a self-administered online survey. In order to research the relationships between variables, an experimental design was used. More specifically, in combination with a survey this is called experimental vignette methodology (EVM) (Blumberg, Cooper, &

Schindler, 2011). EVM can result in important knowledge and insights about causal relationships (Aguinis & Bradley, 2014). In vignette studies, the researcher presents the participants with a description of a situation and then asks the participants to make a decision or evaluate the situation (Blumberg et al., 2011). According to Aguinis and Bradley (2014) dependent variables such as intentions, attitudes, and behaviours can be measured when participants are exposed to realistic scenarios (Aguinis & Bradley, 2014). The dependent variable in this research is intention to book a hotel room and therefore experimental vignette

is suitable.

4.1.1 The experiment

The experimental design used was a 2x2 between-subjects factorial design. Participants were randomly assigned to one of the four conditions, so each participant was exposed to only one condition. Four scenarios were prepared around an unbranded hotel. An unbranded hotel was chosen since Vermeulen and Seegers (2008) found that the persuasive effect was stronger for lesser-known hotels. In other words, well-known hotels can be more resilient to online review effects than lesser-known hotels. To avoid bias from previous experiences of participants, they were told to imagine they are searching for a hotel in an unknown location without previous experience (Mauri & Minazzi, 2013) for a weekend trip. Although the structure of the review page was similar to existing review websites, the main recognition points (such as a name and logo of the existing website) were not included to avoid the influence of brand image (Mauri & Minazzi, 2013). Further information about the hotel, such as price, room facilities and hotel services were also excluded to avoid these to influence hotel room booking intentions of participants. Reviews were created by reading real reviews on review websites

(20)

Methodology | 20 and by amending reviews that are included in studies concerning online reviews (Ladhari & Michaud, 2015; Mauri & Minazzi, 2013). Figure 2shows the information and instruction of the experiment that participants read.

Figure 2: Information of experiment

After reading the message, the participants were exposed to one of the scenarios. For all scenarios participants saw the name of the hotel, a picture of a room and the rating of the hotel.Participants were presented to neutral or negative reviews and there were few or many reviews. More specifically, neutral reviews had a rating of 2.5 (out of five) and negative reviews had a rating of 1 (out of five). Neutral reviews were not very positive nor very negative and words like ‘okay’, ‘reasonable’, ‘basic’, ‘fine’ indicated that reviews were not very good, but not bad either. Few reviews consisted of three reviews and many reviews consisted of ten reviews. The four conditions are shown in the Table 1.

Table 1: 2x2 between subjects factorial design

Volume

Few Many

Valence

Neutral Few neutral reviews (3) Many neutral reviews (10) Negative Few negative reviews (3) Many negative reviews (10)

In condition one participants were exposed to many negative reviews, in condition two participants saw few negative reviews, condition three consisted of many neutral reviews and condition four existed of few neutral reviews. An example of a condition is shown in Figure 3.

(21)

Methodology | 21

Figure 3: Example of condition

After reading the reviews participants answered some questions regarding their intention to book the hotel.

4.2 Measurement of variables

4.2.1 Booking intentions – dependent variable

To measure booking intentions the scale developed by Dodds, Monroe, and Grewal (1991) was used. This is originally a five item scale regarding willingness to buy a product with a Cronbach’s alpha of .97 (Dodds et al., 1991). However, two of these items are about price which was not included in this experiment design. Three items of the scale were used and were about booking a hotel instead of purchasing a product. The three items include

likelihood, probability and willingness to book. An example of an item is: “the likelihood of booking this hotel is very high”. These items are also used by Tsao et al. (2015) and have a Cronbach’s Alpha of .971. This can be measured on a 7 point Likert scale from 1 (strongly disagree) to 7 (strongly agree).

4.2.2 Involvement – control variables

Most factors that could influence hotel room booking intentions such as price, hotel services and room facilities are excluded from the survey. However, not every factor could be

(22)

Methodology | 22 excluded: people might find several elements important besides valence and volume of online reviews. This can be for example the picture of the room, the rating of the hotel, the title of the review or the content of the review. If a person is highly involved with the picture for example, this might result in different hotel room booking intentions than when this person was exposed to online reviews without a picture. To control for this and be sure these factors didn’t interfere with the results, involvement is included in the model as a covariate. To measure involvement of these factors, the personal involvement scale by Zaichkowsky (1994) was used. This is a 10 item scale that is context free. An example of an item is

boring/interesting, the scale ranges from 10 (low involvement) to 70 (high involvement). Cronbach’s Alpha for this scale ranged from .91 to .95 for advertisements and .94 to .96 for products (Zaichkowsky, 1994). This scale was used for the picture of the hotel, the rating of the hotel, the title of the review and the content of the review.

4.2.3 Uncertainty avoidance – moderating variable

In order to measure uncertainty avoidance the CVSCALE – Individual Cultural Values Scale - developed by Yoo et al. (2011) was used. The CVSCALE was developed when a demand aroused for a metric that could measure Hofstede’s cultural dimensions at the individual level (Yoo et al., 2011). It contains 26 items to assess every dimension that Hofstede developed: Power distance; uncertainty avoidance; collectivism; long-term orientation and masculinity.

Since this research focuses only on uncertainty avoidance it was only necessary to use the items that belong to the uncertainty avoidance section, which consists of five items. An example of an item is: “It is important to have instructions spelled out in detail so that I always know what I’m expected to do.” The items can be measured on a 5 point Likert scale from 1 (strongly disagree) to 5 (strongly agree). According to Yoo et al. (2011) the scale is reliable, shows validity and across-sample and across-national generalizability. Prasongsukarn (2009) validated the scale in Thailand and found a Cronbach’s Alpha of .81 for uncertainty avoidance. Furthermore Schumann et al. (2010) found a Cronbach’s Alpha of .86. So, scale reliability is sufficient.

4.3 Survey

The survey was conducted in Qualtrics. Most people in the Netherlands have an adequate level of English and to avoid translation bias the survey was only distributed in English. No translation was necessary. The survey was pre-tested with approximately 8 persons to ensure

(23)

Methodology | 23 that the set of reviews are correctly perceived as either neutral or negative, the questions are understood and that is manipulated what is supposed to be manipulated. After the pre-test the survey was slightly adapted (question wording) and then distributed through Facebook. The order of the survey is the following. Participants first answered some general questions about their experience in booking hotel rooms online and reading reviews. Then the participants were exposed to one of the four conditions and thereafter they answered questions regarding their intention to book the hotel. Next, the items that measure involvement and uncertainty avoidance were asked. Finally some demographic questions (age, gender, nationality) were asked. The survey can be viewed in Appendix 8.2.

(24)

Results | 24

5. Results

To perform the statistical analyses, SPSS software was used. This chapter describes the statistical analysis and the results of this research. First, some information about the sample is given and it is explained how the data is prepared for analysis. Next, the reliability and correlations are described and means are compared. After this, the hypotheses are tested with an ANCOVA analysis for direct effects interaction effects and with a regression analysis for the moderating effects.

5.1 Sample

The sampling technique used is non-probability sampling, specifically convenience sampling. There were no criteria for participants of this research and therefore convenience sampling was a sufficient method.The survey was distributed via social media and through friends and family. A gift card was raffled as an incentive for people to participate in the survey.

The survey ran from the 24th of April until the 14th of May. In total 225 participants fully completed the survey. Most of these participants were Dutch (95.1%) and the majority of participants was female (72%) Mean age of participants is 29 (SD=11.15). Most

participants indicated that when they book a hotel they do this online, 96% of all participants indicated that they booked online almost always or always. Also, most participants are used to reading online reviews. 80.5% of all participants indicated that they almost always or always read online reviews before booking a hotel. The mean score of uncertainty avoidance is 3.68 (SD=.58), which indicates most participants have a moderate level of uncertainty avoidance. The involvement scales show that people are most involved with the content of the review and the rating of the hotel, frequencies show a mean of 5.25 and 5.23 subsequently (1 would indicate low involvement, 7 high involvement). Involvement of the picture and title of the review was slightly lower. Frequencies show a mean of 4.65 and 4.50 subsequently. Table 2 shows the demographics of the participants who completed the survey.

(25)

The influence of e-WOM on hotel room booking intentions Results | 25 Table 2: Demographics (N=225) N % Gender Male 63 28.0 Female 162 72.0 Nationality Dutch 214 95.1 Other 11 4.9

Age (in years)

10-19 3 1.3 20-29 162 72.0 30-39 18 8.0 40-49 19 8.5 50-59 20 8.9 60-69 3 1.3

5.2 Preparing data for analysis

After removal of incomplete surveys, 226 complete surveys remained. Reversed items of the involvement scales were recoded. Means were computed for every variable and a check for outliers was performed. This was done by standardizing scores of variables (possible outliers are cases with z>3). One outlier was found (the uncertainty avoidance scale included an item with z = -3.48) and this item was excluded from further analysis. Furthermore, dummies were computed for the independent variables: valence and volume. After preparing the data for analysis 225 cases remained. Condition one (many negative reviews) exists of 57 surveys, condition two (few negative reviews) exists of 55 surveys, condition three (many neutral reviews) exists of 53 surveys and condition four (few neutral reviews) exists of 60 surveys. 5.3 Reliability

Reliability for the scales booking intentions, uncertainty avoidance and involvement was checked. The booking intentions scale has high reliability, with Cronbach’s Alpha = .940. The corrected item-total correlations indicate that all the items have a good correlation with the total score of the scale (all above .30). Also, none of the items would substantially affect reliability if they were deleted.

The uncertainty avoidance scale has good reliability, with Cronbach’s Alpha = .746. The corrected item-total correlations indicate that all the items have a good correlation with

(26)

Results | 26 the total score of the scale (all above .30). Also, none of the items would substantially affect reliability if they were deleted.

The involvement scale has good reliability, with Cronbach’s Alpha = .910 (the four involvement scales taken together, see Table 5 for separate reliabilities). The corrected item-total correlations indicate that all the items have a good correlation with the item-total score of the scale (all above .30). Also, none of the items would substantially affect reliability if they were deleted.

5.4 Comparing means

To ensure that the groups of the independent variables are different in hotel room booking intentions an independent samples T-Test was conducted for valence and volume. For valence there was a significant difference in the scores for neutral reviews (M= 3.94 SD=1.28) and negative reviews (M=1.88, SD=.99) conditions; t (212.67) =13.57, p=.000. This suggests that valence does have an effect on hotel room booking intentions. Specifically that when people are exposed to neutral reviews, they have higher hotel room booking intentions than when people are exposed to negative reviews. Although significantly higher than negative online reviews, neutral online reviews do not lead to very high hotel room booking intentions, the mean is 3.94 (out of 7) which is slightly above intermediate hotel room booking intentions. Table 3 shows the means of valence.

Table 3: Mean scores of valence

Mean Standard deviation

Neutral 3.93 1.27

Negative 1.88 .99

For volume there was not a significant difference in the scores for few reviews (M=3.09, SD=1.56) and many reviews (M=2.74, SD=1.51) conditions; t (224) =1.71, p=.088. This suggests that volume does not have an effect on hotel room booking intentions. Specifically, when people are exposed to few reviews, they do not have significantly different hotel room booking intentions than when people are exposed to many reviews. The means of volume are reported in Table 4.

(27)

Results | 27

Table 4: Mean scores of volume

Mean Standard deviation

Few 3.07 1.55

Many 2.74 1.51

5.5 Correlations

An overview of the means, standard deviations, correlations and scale reliabilities is presented in Table 5. A few significant correlations can be observed. For example uncertainty avoidance and booking intentions are significant positively correlated. However, the effect is only .14 which is small. One large effect can be observed, valence is significant negatively correlated with booking intentions (-.67). There was no significant correlation found between volume and booking intentions. Furthermore, the different involvement scales are significantly correlated with each other. For example involvement for title of the review and involvement for rating have a correlation of .35. Another large correlation is that between involvement for content of the review and involvement for rating (.48).

(28)

Results | 28

Table 5: Means, Standard Deviations, Correlations and Reliabilities

Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Gender 1.72 .45 -

2. Age 29.12 11.14 -.30** -

3. Nationality 1.05 .22 .00 -.08 -

4. Read online reviews 4.19 .82 -.73 .01 .02 -

5. Book online 4.5 .58 .01 -.13 .05 .22** - 6. Valence 1.50 .50 .08 .04 -.02 -.06 .02 - 7. Volume 1.49 .50 .05 -.02 .07 .01 .02 .04 - 8. Booking intentions 2.92 1.54 -.10 -.04 -.01 -.08 -.10 -.67** -.11 (.94) 9. Uncertainty avoidance 3.67 .59 .03 .03 .09 .10 .02 -.06 -.06 .14* (.75) 10. Involvement Picture 4.65 .95 .10 -.10 .21** .01 .09 -.07 -.01 .13* .17* (.88)

11. Involvement Title of review 4.51 1.08 .05 .04 -.03 .17* .17* .03 -.20** -.11 .01 .20** (.92)

12. Involvement Content of review 5.25 .81 .08 -.09 .11 .22** .13* -.06 -.01 -.06 .12 .21** .27** (.85)

13. Involvement Rating of hotel 5.23 .80 -.01 -.11 .04 .21** .17* -.02 .03 -.13 .06 .22** .35** .48** (.84) ** Correlation is significant at the 0.01 level (2-tailed)

(29)

Results | 29 5.6 Testing hypotheses

5.6.1 Direct effects & interaction effect

A factorial ANCOVA (analysis of covariance) was computed to examine the main effects of valence and volume on hotel room booking intentions and to test if there was an interaction effect between valence and volume, while controlling for involvement (involvement picture, involvement title of review, involvement content of review, involvement rating). This

research is a 2x2 between subjects experiment and a factorial ANOVA (analysis of variance) is a suitable model to analyse main effects and interaction effects. When covariates are measured and included in the ANOVA model, it is called an ANCOVA (Field, 2014).

The independent variables both include two levels, valence consists of neutral reviews and negative reviews and volume consists of few reviews and many reviews. Group sizes are not entirely equal. For valence, 113 participants were exposed to neutral reviews and 112 to negative reviews. For volume, 115 participants were exposed to few reviews and 110 to many reviews.

There are several assumptions for an ANCOVA analysis. The first is homogeneity of variance. Levene’s test reports p<.05, so the null hypothesis that assumes equal variances needs to be rejected. According to Field (2014) transforming the data can rectify the problem of unequal variances. The dependent variable booking intentions is logarithmic transformed (ln) to correct this issue and after transformation Levene’s test reports p>.05, so the null hypothesis that assumes equal variances is not rejected. Consequently, equal variances can be assumed. Another assumption of ANCOVA is independence of the covariate and treatment effect (Field, 2014). This should not be a problem when participants are randomly assigned to an experimental group (Field, 2014) but is measured with ANOVA (Appendix 8.3.1) to be certain this assumption is not violated. The assumption is violated for one covariate, involvement title of review is significantly different for the independent variable volume. However, since this assumption is not violated for any other variable and Field (2014) writes that ANCOVA should be unbiased when groups differ on levels of a covariate, it is expected that it is not an issue. Another assumption is homogeneity of regression slopes, to test this an ANCOVA was performed by using a customized model (Field, 2014) (Appendix 8.3.2). The assumption of homogeneity of regression slopes has been broken for one of the control variables, the relationship for involvement of the picture is different for the groups of volume.

(30)

Results | 30 Since the assumption is not violated for other variables, ANCOVA seems suitable and the analysis is performed. Table 6 shows the results of the factorial ANCOVA.

Table 6: Factorial ANCOVA

SS DF MS F η2 Sig.

Involvement picture 1.07 1 1.07 6.85 .03 .01

Involvement title of review .03 1 .03 .18 .00 .67

Involvement content of review .56 1 .56 3.55 .02 .06

Involvement rating .97 1 .97 6.20 .03 .01 Valence 34.55 1 34.55 220.82 .50 .000 Volume .49 1 .49 3.12 .02 .08 Valence*Volume .38 1 .38 2.46 .01 .12 Error 33.95 217 .16 Total 261.58 225

Significant at the p<.05 level

The first covariate, involvement picture, was significantly related to booking intentions, F (1, 217) = 6.85, p <.01, η2 = .03. The second covariate, involvement title of review was not significantly related to booking intentions, F (1,217) = .18, p>.05, η2_{=.00. The third covariate,} involvement content of review, was not significantly related to booking intentions, F (1,217) = 3.55, p>.05, η2_{=.02. The fourth variable, involvement rating, was significantly related to} booking intentions, F (1, 217) = 6.2, p<.05, η2_{=.03. Two out of four control variables were} significant and therefore it was right to include these in the model. Apparently, there was an effect of these covariates on the dependent variable, now this is controlled for in the model.

A significant main effect of valence on booking intentions was found, after controlling for the effect of involvement, F (1, 217) = 220.82, p < .01, η2=.50. There was a

non-significant main effect of volume on booking intentions, after controlling for the effect of involvement, F (1,217) = 3.12, p>.05, η2=.01. Furthermore, there was a non-significant interaction effect between Valence and Volume on booking intentions, after controlling for the effect of involvement, F (1, 217) =2.46, p>.05, η2

(31)

Results | 31

Figure 4: Plot factorial ANCOVA

(32)

Results | 32 For valence (p<.01) the null hypotheses can be rejected, which means there is a difference between the two groups of valence (neutral/negative reviews) and there is a direct effect on booking intentions. In Figure 4 and Figure 5 it can be observed that neutral online reviews lead to higher booking intentions than negative reviews. So, H1 can be supported, there is a direct effect of valence on booking intentions and the effect is more positive for neutral reviews than for negative reviews.

For volume (p>.05) the null hypothesis cannot be rejected, which means there is no difference between the groups of volume (few/many reviews) and there is no direct effect on booking intentions. In Figure 4 and Figure 5 it can be observed that many reviews lead to slightly lower booking intentions than few reviews but this is not a significant difference. Therefore H2 cannot be supported, there is no direct effect of volume on booking intentions.

For the interaction between valence and volume (p>.05) the null hypothesis cannot be rejected, which means there is no interaction effect between valence and volume. Therefore, H3 cannot be supported, the effect of valence on booking intentions is not strengthened significantly with a higher volume of reviews.

5.6.2 Moderation

Process was used to run the moderation in SPSS. Model one was used since the model is a simple moderation. The moderator uncertainty avoidance is numerical (interval; Likert scale) and is therefore standardized in order to use Process correctly. In Table 7 and Table 8 the results for the moderation on the effect of valence on booking intentions can be found.

Table 7: Moderation on the effect of valence on booking intentions

Coefficient SE t p Intercept i1 _2.10 _.08 _25.67 _.000 Valence (X) c1 _-.79 _.06 _-14.07 _.000 Uncertainty avoidance (M) c2 _.05 _.10 _.47 _.640 Valence*Uncertainty avoidance (XM) c3 _.00 _.07 _.06 _.953 r2 = .49 p<0,001 F(3,222)=68.07

(33)

Results | 33

Table 8: Moderation on the effect of valence on booking intentions (2)

Conditional effect of Valence (X) on Booking intentions (Y) at levels of Uncertainty avoidance (M)

Uncertainty avoidance Effect SE t p

-1,46 -.79 .11 -7.21 .000

-.79 -.79 .08 -10.50 .000

.22 -.79 .06 -13,52 .000

.56 -.79 .07 -11.60 .000

.89 -.79 .08 -9.53 .000

The regression coefficient for XM is c3= .00 and is statistically not different from zero, t (222) = .06, p = .95. Thus, the effect of valence on booking intentions does not depend on the level of uncertainty of a person. Since p>.05 there is no moderating effect of uncertainty avoidance on the effect of valence on booking intentions. Therefore, H4a is not supported. In Table 9 and Table 10 the results for the moderation on the effect of volume on booking intentions can be found.

Table 9: Moderation on the effect of volume on booking intentions

Coefficient SE t p Intercept i1 _1.12 _.12 _9.44 _.000 Volume (X) c1 _-.13 _.08 _-1.74 _.084 Uncertainty avoidance (M) c2 _-.03 _.13 _-.23 _.818 Valence*Uncertainty avoidance (XM) c3 _.07 _.08 _.88 _.381 r2 = .03 p=04. F(3,222)=2,74

Table 10: Moderation on the effect of volume on booking intentions (2)

Conditional effect of Volume (X) on Booking intentions (Y) at levels of Uncertainty avoidance (M)

Uncertainty avoidance Effect SE t p

-1.46 -.24 .14 -1.69 .092

-.79 -.19 .10 -.1.91 .057

.22 -.12 .08 -1.48 .142

.56 -.09 .09 -1.02 .308

(34)

Results | 34 The regression coefficient for XM is c3= .07 and is statistically not different from zero, t (222) = .88, p = .38. Thus, the effect of volume on booking intentions does not depend on the level of uncertainty of a person. Since p>.05 there is no moderating effect of uncertainty avoidance on the effect of volume on booking intentions. Therefore, H4b is not supported.

5.7 Summary of results

In Table 11 the hypotheses are listed and a summary of the results is provided.

Table 11: Summary of results

H1 Valence of online reviews has a direct effect on hotel room

booking intentions. The effect is more positive for neutral reviews than for negative reviews

Supported

H2 Volume of online reviews has a direct effect on hotel room booking intentions. The effect is more positive for many reviews than for few reviews.

Not supported

H3 There is an interaction effect between valence and volume on hotel room booking intentions. The effect of valence on hotel room booking intentions is strengthened with a higher volume of reviews.

Not supported

H4a The effect of valence of online reviews on hotel room booking intentions is moderated by individual level uncertainty avoidance. The effect of valence of online reviews on hotel room booking intentions is stronger for people high in uncertainty avoidance than for people low in uncertainty avoidance

Not supported

H4b The effect of volume of online reviews on hotel room booking intentions is moderated by individual level uncertainty avoidance. The effect of volume of online reviews on hotel room booking intentions is stronger for people high in uncertainty avoidance than for people low in uncertainty avoidance.

(35)

Discussion | 35

6. Discussion

In this chapter the findings are discussed, together with the theoretical and practical

implications of these findings. Additionally, the limitations of this research and suggestions for future research are provided.

6.1 Main findings

This research looked into the more and more researched domain of online reviews in the hotel industry. Multiple factors of online reviews can have an effect on how these reviews are perceived and this can influence awareness, attitudes and booking intentions. A 2x2 between-subjects factorial design was executed to research the effect of valence and volume of online reviews on hotel room booking intentions. It was hypothesised that both online review

valence and volume have a direct effect on intentions to book a hotel room. Valence consisted of neutral and negative reviews and volume consisted of few and many reviews, which was 3 and 10 reviews respectively. It was hypothesised that the effect of online reviews on booking intentions is more positive for neutral reviews than for negative online reviews and that the effect is more positive for many than for few online reviews. An interaction effect was also expected. The hypothesis was that the effect of valence on booking intentions was

strengthened by a higher volume of reviews.

Individual culture was expected to have an influence on how online reviews are perceived. A cultural factor that is likely to have an influence is uncertainty avoidance. It is found that online reviews reduce risk and uncertainty (Kim et al., 2011) and therefore people that are high in uncertainty avoidance are more influenced by online reviews and search for more information than people low in uncertainty avoidance. It was hypothesised that the effect of valence and volume of online reviews on booking intentions is stronger for people high in uncertainty avoidance than for people low in uncertainty avoidance.

In line with research of Ladhari and Michaud (2015); Mauri and Minazzi (2013); Sparks and Browning (2011), Tsao et al. (2015) and Ye et al. (2011) which all look at the valence of reviews and booking intentions, this research found a significant effect of valence on booking intentions. Negative reviews lead to very low booking intentions, which is in agreement with the negativity effect: people place greater weight on negative information (Skowronski & Carlston, 1989). In accordance with Mudambi and Schuff (2010) who found that for experience goods moderate reviews are more helpful and credible than extreme reviews, this research found similar results. Neutral reviews lead to significantly higher

(36)

Discussion | 36 booking intentions than negative reviews do. This accords the extremity effect, which

explains that people put more weight on extreme reviews than on moderate reviews (Skowronski & Carlston, 1989). Neutral reviews led to booking intentions slightly above moderate booking intentions. This means that neutral reviews do not lead to very high or low booking intentions: it is not very enhancing and also not detrimental to booking intentions. Thus, the first hypothesis was supported, neutral reviews lead to higher booking intentions than negative reviews.

Although a direct relationship between volume and booking intentions can be expected because of the awareness effect (Chen, Wang & Xie, 2011; Liu, 2006; Park et al., 2007; Tsao et al., 2015) and the consideration set framework (Roberts & Lattin, 1991) this study found no significant direct effect of online review volume on booking intentions. It does not significantly matter if there are few or many reviews. An interaction effect between

valence and volume was not found either, although previous research by Tsao et al. (2015) about valence and volume found that a high number of positive reviews received higher booking intentions than a low number of positive reviews. A high number of negative reviews received lower booking intentions than a low number of negative reviews. Although it can be observed that many negative reviews, rather than few negative reviews, lead to lower booking intentions, but this effect is not significant. Further, for neutral reviews this cannot be

observed.

There might be several reasons for the non-significant findings. The online reviews were written in English while most participants were Dutch (95.1%). When Dutch people are looking for reviews they will probably read those reviews in Dutch. Providing the online reviews in the native language of participants might make the reviews more credible. Since Dutch participants would probably visit the Dutch version of a review website the set of reviews provided to them now might not have been believable enough. Also, volume was probably not manipulated well enough, the independent samples T-Test showed that groups of volume did not significantly differ in hotel room booking intentions. These factors might have influenced the results.

For uncertainty avoidance no significant moderating effect was found for valence and volume on booking intentions. For valence, this was not in line with the research of

Schumann et al. (2010) who mention that people high in uncertainty avoidance are more susceptible to reviews. For volume, a moderating effect was expected since people from high

(37)

Discussion | 37 uncertainty avoidance cultures read more information before making a decision than people from low uncertainty avoidance cultures (Money, Gilly, & Graham, 1998). The mean

uncertainty score was 3.68 (out of 7). This indicates that most people were not extremely low or high in uncertainty avoidance. The non-significant finding might be due to this lack of variance of levels of uncertainty avoidance among participants.

6.2 Theoretical and practical implications

This research contributed to research by examining the effect of online reviews in the hotel business, especially by looking into the effect of neutral online reviews and the influence of individual cultural differences. Whereas the majority of studies looked at the effect of positive and negative online reviews on booking intentions (e.g. Ladhari & Michaud 2015, Mauri & Minazzi, 2013; Tsao et al., 2015), this research proposes another type of valence, neutral online reviews. Online reviews are not always truly positive or negative and therefore researching neutral online reviews sheds new light into the topic of online reviews in the tourism and hospitality industry.

Most research into cultural differences considers cross-cultural differences, this research looked at individual cultural differences. Researching culture as a factor that can influence the effect of online reviews responds to Mauri and Minazzi (2013), suggestion more research into how behaviours of people of different cultures are effected by e-WOM. This research also adds to the study of Schuman et al. (2010) by researching the effect of negative WOM and the moderating effect of uncertainty avoidance in the service industry.

In practice, this research has implications for the tourism and hospitality industry, in particular hotel managers might benefit from the results of this research. It raises awareness about the importance of online reviews and the possible effects on the financial performance of their hotel (Ladhari & Michaud, 2015). Based on the findings concerning the valence of online reviews, hotel managers should try to reduce negative online reviews (Mauri & Minazzi, 2013). Additionially, hotel managers should not stimulate nor try to reduce neutral reviews. Neutral reviews are not damaging to booking intentions (they lead to slightly positive booking intentions) and variance in opinions of online reviews does not influence booking intentions (Ye et al., 2011). Hence, having some neutral reviews in the set of total reviews does not harm booking intentions and might increase credibility (Mudambi & Schuff, 2010). Trying to reduce negative reviews can be done in several ways: improving the service quality and service recovery while the customer is still staying at the hotel and stimulating

The influence of e-WOM on hotel room booking intentions : An investigation on the effect of valence and volume & the moderating effect of individual level uncertainty avoidance