Why and when do tourists
share photos on social media?
A case study for Amsterdam city.University of Amsterdam
Faculty of Economics and Business
Master of Science in Business Administration Track: Marketing
Student: Marije van Oostenbruggen Student number: 6177077
Under supervision of: Bob Rietveld Date: 24 June 2016
Statement of originality
This document is written by Student Marije van Oostenbruggen 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
ACKNOWLEDGEMENTS
This research was partly supported by Amsterdam Marketing. I would like to thank Olivier Ponti, research manager at Amsterdam Marketing, for the inspiring cooperation and sharing valuable knowledge and insights.
I would also like to thank my supervisor Bob Rietveld for his guidance and support throughout this research and his helpful comments.
TABLE OF CONTENTS
INDEX OF FIGURES...6
INDEX OF TABLES...6
ABSTRACT...7
1. INTRODUCTION...7
1.1 Introduction...7
1.2 Purpose of the study...9
1.3 Relevance...9
2. THEORETICAL FOUNDATION AND RESEARCH HYPOTHESES...11
2.1 Peak-‐end theory...11
2.1.1 Peak-‐end theory in holiday setting...11
2.1.2 Timing...13
2.2 Drivers of eWOM...13
2.2.1 Utility and accessibility...13
2.2.2 Emotions...14
2.2.3 Self-‐enhancement...14
2.3 Overall satisfaction...15
2.4 Sharing motivations...15
2.4.1 Arousal related sharing motivations...15
2.4.2 Community-‐ versus self-‐related sharing motivations...16
3. METHOD...18 3.1 Introduction...18 3.2 Sample description...18 3.3 Pre-‐test...20 3.4 Measurements...20 3.5 Study design...25 3.6 Data collection...26 3.7 Data analysis...26 4. RESULTS...27
4.1 Treatment of missing data...27
4.2.1 Peak-‐end theory...27
4.2.2 Social media...31
4.2.3 Emotions & sharing behavior...32
4.2.4 Overall satisfaction...34
4.2.5 Sharing motivations...36
4.4 Summary of hypothesis…...39
5. CONCLUSION & DISCUSSION...40
5.1 Conclusion & discussion...40
5.2 Limitations & future research...42
REFERENCES...43
APPENDIX A. Questionnaire...48
INDEX OF FIGURES
Figure 1. Conceptual model...12
Figure 2. Circular ordering of affect descriptors (Russell & Pratt, 1980)...21
Figure 3. Distribution peak moment 1...28
Figure 4. Distribution peak moment 2...28
Figure 5. Data visualization of photo-‐making and photo-‐sharing...30
INDEX OF TABLES Table 1. Research gap for present study...8
Table 2. Nationalities of the sample...19
Table 3. Length of stay of the respondents...19
Table 4. Emotions constructs...23
Table 5. Rotated component Matrix emotions...22
Table 6. Rotated component Matrix sharing motivations...25
Table 7. Social media use by everyone (N = 113)...31
Table 8. Social media use by 25 or younger (N = 56)...32
Table 9. Social media use by 26 or older (N = 57)...32
Table 10. Timing of sharing peak moments...32
Table 11. Correlation table overall satisfaction and emotions...35
Table 12. Regression analysis overall satisfaction and emotions...35
Table 13. Correlation table overall satisfaction and sharing motivations...38
ABSTRACT
While the amount of visual content shared by tourists on social media is growing, little is known about what and why tourists share visual content on social media. This study contributes to the understanding of the social media behavior of tourists by zooming in on the two peak moments of a holiday experience. An en route survey was conducted among 121 tourists who just visited Amsterdam. The results show that approximately half of the peak moments were shared on social media. Peak moments that were shared on social media scored higher on positive and high arousal emotions than those peak moments that were not shared. The results also show that tourists are more driven by community-‐related than self-‐related motivations to share their peak moments. This study shows that analysing peak moments provides valuable insights for marketing practitioners. However, further research is required in order to gain more insights into what determines whether people share peak moments or not.
Keywords: Peak-‐end theory, social media, visual eWOM, travel-‐experience sharing
1. INTRODUCTION 1.1 Introduction
How does a holiday experience influence the attitude and behavior of tourists? The holiday experience has been a major research topic for several decades. Research can be divided into two broad streams, of which the first stream has focused on the effect of holiday experiences on attitude, which includes subjects as overall satisfaction (Wang, Park & Fesenmaier, 2012), recalled emotions (Nawijn, 2010; Nawijn, Mitas, Lin & Kerstetter, 2013) and destination image formation (Kim & Chen, 2015; Ekinci & Hosany, 2006). The second stream that can be distinguished is focused on the effect of holiday experiences on behavior. Researchers for example investigated whether tourists (intended to) return to the same destination (Alia, Hussain & Ragavan, 2014; Bigné, Sánchez & Sánchez, 2001), but the most dominant subject within this stream is the behavior of sharing holiday experiences with others. Until about ten years ago, people who returned from their holiday shared their holiday experiences by telling stories, and maybe even more important: photographs. By using a slideshow or photo book, people showed these photographs to a limited amount of people. The rise of internet, the emergence of many different social media platforms and the continuous improvement of smartphones and digital cameras, enables people to share visual content with many others. Perhaps the selection of holiday photos being shared has shrunk,
but the amount of people reached has significantly increased. Therefore electronic visual word-‐of-‐ mouth (eWom) is an emerging area of interest. Munar and Jacobsen (2014) already found dominance in sharing visual content versus narrative content when sharing holiday experiences online. Munar and Jacobsen suggest that sharing information is more related to textual communicative practices, and sharing experiences is more suited for visual content. When evaluating affective experiences, such as a holiday, Fredrickson (2000) states that there are only two important moments that come to mind easily, which Fredrickson calls the Peak-‐end theory. Fredrickson states that these two moments guide people’s choices in which experiences to avoid, repeat and recommend to others (WOM). Could this implicate that the Peak-‐end theory also explains which moments people share on social media (eWOM)? Several studies attempted to apply the Peak-‐end theory within the tourism context and investigated whether the Peak-‐end theory can predict attitudes, such as evaluations and experienced emotions (Kemp, Burt & Furneux, 2008; Geng, Chen, Lam & Zheng, 2013). However, no research was found that tested whether the Peak-‐end theory can also predict behavior (Table 1). By examining this assumption, this study sets the first steps for future research, for using the Peak-‐end theory as guidance. In this research, a survey was designed to attempt to relate the Peak-‐end theory to visual content sharing behavior of tourists on social media.
Independent variable
General holiday experience Peak-‐end theory
Dep en den t va ria bl e
Attitude Ekinci & Hosany (2006) Nawijn (2010)
Wang, Park & Fesenmaier (2012) Nawijn, Mitas, Lin & Kerstetter (2013) Kim & Chen (2015)
Fredrickson (2000)
Kemp, Burt & Furneux (2008) Geng, Chen, Lam & Zheng (2013)
Behavior Huang, Basu & Hsu (2010) Kang & Schuett (2013) Munar & Jacobsen (2014)
Present study
Table 1. Research gap for present study
1.2 Purpose of the study
The main purpose of this research was to provide more insights into the use of social media by tourists for sharing holiday experiences, by zooming in on the two peak moments of a holiday. The main question was:
“Why and when do tourists share photos on social media?”
To answer this main question, the following research questions were answered: 1. Do people share peak moments on social media?
2. Why do people share peak moments on social media?
3. Which social media platforms do tourists use to share peak moments? 4. When are peak moments shared on social media?
5. What emotions do tourists experience during peak moments and how do these emotions affect sharing behavior on social media?
6. Do age, gender, holiday purpose or length of stay explain variation in sharing behavior of peak moments?
1.3 Relevance
The importance of social media in marketing is acknowledged by an increasing number of industries, including tourism (Munar & Jacobsen, 2014; Zehrer & Grabmüller, 2012). The Amsterdam Visitor Survey 2012, that is held every five years among more than 10,000 tourists visiting Amsterdam, shows that Word-‐of-‐mouth (WOM) is the most consulted information source for new visitors before making the decision to visit Amsterdam. WOM is perceived as more credible and trustworthy than traditional marketing activities (Cox, Burgess, Sellitto & Buultjens, 2009) and affects both attitude and behavior of the recipient (Trusov, Bucklin & Pauwels, 2009; Zarrad & Debabi, 2015). Zeng and Gerritsen (2014) found that eWOM significantly contributes to the reputation of destinations and that eWOM plays a role in the entire travel cycle: before, during and after a trip. Numerous researchers have studied textual eWOM. Visual eWOM, however, has not received much attention yet (King, Racherla & Bush, 2014), apart from several researchers (Munar & Jacobsen, 2014; Ring, Tkaczynski & Dolnicar, 2014; Haldrup & Larsen, 2003) who showed a constant increase in sharing visual content online. This research provides more insights into the sharing behavior and underlying motivations of visual eWOM by tourists. While other studies (Huang, Basu & Hsu, 2010; Munar & Jacobsen, 2014) investigated the intention to share travel experiences, this study investigates the actual sharing behavior. A better understanding of this
behavior is helpful for both the travel industry and other service industries to identify which elements influence the behavior of people sharing visual content online. To be more specific, this research is focused on social media. Nowadays more than two billion people worldwide are in possession of a smartphone (emerce.nl, 2015). By the increase of free wifi-‐spots and a decrease in costs of internet data, people have access to social media almost everywhere and anytime. The instant access to information obtaining and sharing influences both the emotional states and behavior of tourists by enabling them to solve problems, share experiences and store memories (Wang, Park & Fesenmaier, 2012). A better understanding of how different social media platforms encourage tourists to instantly store and share their holiday experiences is essential according to Kozinets, de Valck, Wojnicki and Wilner (2010).
Another way in which this research contributes to the current knowledge is by specifically focusing on two peak moments. Most studies that have examined sharing behavior of tourists investigated whether tourists shared or intended to share information about their trip in general (Munar & Jacobsen, 2014; Huang, Basu & Hsu, 2010). A trip in general is a very broad concept, which leaves much room for speculation about what kind of information people actual shared. The Peak-‐end theory seemed a suitable framework for creating a questionnaire that is more specific about what kind of information tourists would actually share. Therefore, this research focuses on the two main peak moments that someone has experienced during a holiday. Until now, no research has been conducted that investigated whether the Peak-‐end theory can serve as a sort of predictor of the kind of information that tourists share on social media. If the Peak-‐end theory is able to explain/predict the moments of which people post photos online, destination-‐marketing organisations (DMO’s) could perform big data content analysis to discover which activities cause peaks and give new directions to future marketing campaigns.
2. THEORETICAL FOUNDATION AND RESEARCH HYPOTHESES
2.1 Peak-‐end theory
2.1.1 Peak-‐end theory in holiday setting
Fredrickson (2000) talks about the evaluation of events in lives, which Fredrickson calls episodes. Examples of episodes are going to the movie and holidays. These episodes include a beginning and an end, and many moments in between. When evaluating these episodes, Fredrickson (2000) states that there are only two important moments that easily come to mind, which Fredrickson calls the Peak-‐end theory. It suggests that the peak and the end of an affective experience matter more in “retrospective evaluations” than all the other moments combined. A peak is defined as “the most intense affective moment” and the end is “the affect experience in the end” (Fredrickson, 2000, p. 585). The duration of the episode itself does hardly contribute to the evaluation because of the ‘attentional phenomenon’ (Fredrickson & Kahneman, 1993, p. 54). Fredrickson and Kahneman (1993) state that people are aware of the duration of an experience, but the salient moments that come to mind most rapidly, are the ones that determine the retrospective evaluation. Several studies have confirmed this theory by various experiments (Fredrickson & Kahneman, 1993; Redelmeier & Kahneman, 1996; Fredrickson, 2000). These two moments guide people’s behavior for the future, for example in return intention and recommendations to others (Fredrickson, 2000). Both behaviors are extremely relevant for the tourism business.
Holidays include a beginning and an end, which makes holidays well suited for testing the Peak-‐end theory (Fredrickson, 2000). Kemp, Burt and Furneaux (2008) attempted to test the Peak-‐end theory in a holiday setting. Kemp, Burt and Furneaux asked respondents during their holiday on a daily basis about their happiness to find prove for the Peak-‐end theory as a predictor for their holiday memories, but the results only supported half of the theory. Kemp, Burt and Furneaux found that recalled happiness could be predicted by the end-‐part, but not by the peak. Kemp, Burt and Furneaux suggested that this had to do with the fact that extreme affection can fade over time. Geng, Chen, Lam and Zheng (2013) support these thoughts, as they found through literature research that the Peak-‐end theory appears to be a good explanation on a short retention interval, rather than over a long retention interval (Geng, et al., 2013, p. 225).
Nawijn (2010) also investigated the level of happiness during a holiday experience, which Nawijn calls the ‘holiday happiness curve’. Nawijn tested the ‘mood’ of tourists by asking the question ‘How are you feeling today?’, using a 10-‐point scale (1 = terrible, 10 = excellent). Nawijn (2010) did find that people experience the highest point of happiness during the core phase, which could be coded as a peak, and end phase of their holiday. By asking tourists to rate their mood at that moment of time, Nawijn used a moment-‐based measurement. In 2000, Kahneman made a distinction between a moment-‐based approach and a memory-‐based approach to measure the experienced utility and emotions of an affective and hedonic experience. According to Kahneman, the moment-‐based approach, on one hand, is most appropriate when measuring “the experienced utility of an episode” (Kahneman, 2000, p. 17). On the other hand, the memory-‐based approach is used to measure the retrospective evaluations on an episode, which makes the memory-‐based approach seem more appropriate to test the Peak-‐end theory. A holiday represents an affective and hedonic experience, and therefore it is expected that the Peak-‐end theory would apply in a holiday setting (Fredrickson 2000; Kemp, Burt & Furneux, 2008). To test whether the Peak-‐end theory exists in a holiday setting, a memory-‐based approach was utilized to test the following theoretical framework (figure 1).
Figure 1. Conceptual model
2.1.2 Timing
Fredrickson (2000) suggests that the “peak” and the “end” of an affect experience determine the retrospective evaluation. What people evaluate as a peak is completely subjective. However, the timing of these moments is not. To test whether one of the two moments took place in the end part of a holiday experience, the first hypothesis states:
Hypothesis 1: One of the two peak moments took place in the end of a holiday experience.
2.2 Drivers of eWOM
The connection between the Peak-‐end theory and holiday experience has been made only by a few researchers (Nawijn, Mitas, Lin & Kerstetter, 2013; Kemp, Burt & Furneaux, 2008), but even more notable is that there is no study found that investigated the connection between the Peak-‐end theory and social media sharing behavior. A peak moment is a very intense affective moment that comes to mind when people evaluate an experience. In the case of a holiday, this would implicate that if people who just visited a destination, were asked about their holiday experience, they would mention the peak and the end moment. These two moments are the moments that people share with their friends and family (WOM). This could implicate that these moments are also shared on social media, since social media influences tourists to share experiences and store memories (Wang, Park & Fesenmaier, 2012). Cappella, Kim and Albarracín (2015) proposed four drivers for transmitting eWOM: utility, accessibility, emotion and self-‐enhancement. The first paragraph explains how utility and accessibility could explain sharing behavior of peak moments, followed by two paragraphs explaining how emotion and self-‐enhancement could possibly predict the sharing behavior of peak moments on social media.
2.2.1 Utility and accessibility
Two of the four factors that drive people to share information with others are utility and accessibility (Cappella, Kim & Albarracín, 2015). By these factors Cappella, Kim and Albarracín meant that when people communicate with each other, both online and offline, they are more likely to talk about things that are on top of their mind (Berger & Iyengar, 2013). As Fredrickson (2000) stated that the Peak-‐end theory can predict retrospective
evaluation based on the two important moments that come to mind easily, these two factors could possibly explain why people would share peak moments on social media.
2.2.2 Emotions
Another factor that explains sharing eWom is emotion. Fredrickson (2000) suggests that the Peak-‐end theory can be explained by the intensity of emotions of having personal meaning. Hosany and Witham (2006) defined a holiday experience as a unique, emotionally charged experience of high personal meaning, which makes a holiday well suited to use the Peak-‐ end theory. Fredrickson (2000) states that an emotion with high personal meaning, such as love or shame, is expected to be dominant in the evaluation of an affective experience. These emotions are accompanied by a high level of arousal. Arousal can refer to internal excitement, stimulation, exhilaration or inspiration (p. 157 in Neal, Sirgy & Usyal, 1999). According to several researchers emotional arousal is identified as a determinant of sharing information (Berger & Milkman, 2012; Berger, 2011; Cappella, Kim & Albarracín, 2015). Berger (2011) suggests that transmission is partly driven by arousal, because arousal is characterized by activation that could boost sharing (p. 891 in Berger, 2011). Also Berger and Milkman (2012) found that content that evokes high arousal emotion is more viral, regardless the valence of the textual content. Since peak moments are often accompanied by a high level of arousal, this research investigated whether high arousal emotions activate people to share their holiday experiences. The following hypothesis was proposed:
Hypothesis 2a: High arousal emotions lead to a higher chance of sharing peak moments than low arousal emotions.
2.2.3 Self-‐enhancement
Based on the human tendency to self-‐enhance, Berger (2014) states that people want to be perceived positively and therefore try to present themselves in a way that supports a positive impression. Berger and Iyengar (2013) investigated how the medium that enables WOM, shapes the message. Berger and Iyengar looked to the differences between oral versus written communication and found that people tend to share more interesting things when they use written communication. Berger and Iyengar state that this effect is also driven by the tendency of self-‐enhancement and the asynchronous character of the medium platform, because when people can write, they have more time to construct and refine the
message. Visual content is a different form of communication. Although it is not possible to construct a photo once the moment has passed, it is possible to carefully select a photo and also edit the photo in order to refine the image. It is for these reasons that sharing visual content allows people to consciously decide what image they send out to others in order to represent themselves. Therefore it is expected that positively valenced information, in this case positive peak moments, is more widely shared (Cappella, Kim & Albarracín, 2015) to enhance a positive image compared to negative peak moments. The following hypothesis is:
Hypothesis 2b: Peak moments that are shared score higher on positive emotions and peak moments that are not shared score higher on negative emotions.
2.3 Overall satisfaction
Customer satisfaction is important for any kind of organization in order to gain and retain customers (Bigne, Sanchez & Janchez, 2001). Therefore many studies already investigated the causes and effects of customer satisfaction. Satisfied customers appear to share more positive WOM (Anderson, 1998; Kang & Schuett, 2013; Kim, Holland & Han, 2013). Within the tourism industry, the (e)WOM of recent tourists can be of great value in the orientation and organisation phase of potential tourists to visit a destination (Xiang & Gretzel, 2010; Nezakati, et al., 2015; Pan, MacLaurin & Crotts, 2007). Therefore it is highly relevant for destination marketing organizations (DMO’s), such as Amsterdam Marketing, to gain more insight into the level of overall satisfaction of tourists, possibly derived from sharing visual eWOM. In the context of the Peak-‐end theory, this could implicate that people who shared peak moments on social media as a form of eWOM are overall more satisfied about their holiday experiences. Therefore the following hypothesis is:
Hypothesis 3: Sharing peak moments is positively related to overall satisfaction.
2.4 Sharing motivations
2.4.1 Arousal related sharing motivations
Berger (2014) made a distinction between motivations for people to contribute to social media. Berger distinguished five motivations: Impression management, Information acquisition, Social bonding, Emotion regulation and Persuading others. According to Berger three of these five motivations are partly driven by arousal: (1) The motivation of Impression management could lead to sharing high arousal content, since this information seems more
interesting, entertaining or engaging for the receiver. Visual content is a substantially easier and faster way to share an impression of an experience with others, compared to textual content. (2) The motivation of Emotion regulation could lead to sharing high arousal content because it allows people to get rid of the excitement aroused feeling. High arousal emotions are associated with a higher level of activation, which encourages sharing behavior as well (p.594 in Berger, 2014). (3) The third motivation is Persuading others, which increases the chance of sharing high arousal content due to the association with a higher level of activation and trying to activate others. It is expected that people who share high arousal peak moments are highly motivated by these three arousal-‐related sharing motivations:
Hypothesis 4a: Arousal-‐related sharing motivations of peak moments are positively related to sharing high arousal peak moments.
2.4.2 Community-‐ versus self-‐related sharing motivations
Munar and Jacobsen (2014) also investigated tourists’ motivations to share travel experiences in general on social media. Munar and Jacobsen grouped two types of sharing motivations, namely: community-‐related motivations and self-‐related motivations. Community-‐related motivations concern expectations about possible impacts of a person’s behavior on others, while self-‐related motivations concern expectations about the impact of a person’s behavior on the person him-‐ or herself (p. 48 in Munar & Jacobsen, 2014). The two community-‐related motivations were (1) ‘I want to inform others’ and (2) ‘I want to encourage others to visit …’. The three self-‐centered motivations were (1) ‘I want to share my impressions’, (2) ‘I want to be recognised for my experiences’ and (3) ‘I want to maintain social connections’. These five motivations can also be subdivided into the five motivations to contribute to social media of Berger (2014). The first is Impression management, which corresponds to ‘I want to share my impressions’. The second motivation is Information acquisition, which corresponds to ‘I want to inform others’. The third motivation is Social bonding, which is measured in this research with the question: ‘I want to maintain social connections’. The fourth motivation is Emotion regulation, which matches with ‘I want to be recognised for my experiences’ and the last motivation is Persuading others, with corresponds to ‘I want to encourage others to visit Amsterdam’.
Munar and Jacobsen found that community-‐related motivations are most relevant for information sharing. Munar and Jacobsen note that motivational factors differ per type of
content and type of social media. This research concentrated on a specific type of content, namely photos of peak moments. Peak moments are hypothesized to positively influence the level of overall satisfaction (hypothesis 3a) and people who are more satisfied, produce more WOM in order to inform others, as they want to actively contribute to online communities (Wang & Fesenmaier, 2003). Therefore it is expected that:
Hypothesis 4b: Sharing peak moments is positively related to community-‐related motivations.
Hypothesis 4c: Community-‐related motivations are positively related to a high level of overall satisfaction.
3. METHOD 3.1 Introduction
This research aimed to gain more information about the sharing behavior of peak moments of a holiday experience. In order to explore the research objectives, a questionnaire was designed. The next chapter describes the design process of the survey, sample and procedure.
3.2 Sample description
The data was collected in two weeks within the month of April 2016. The target population included all tourists who had just visited Amsterdam, for either holiday or business purpose and travelled by train on their way to the Airport. Respondents had to meet three criteria, namely: (1) they had just visited Amsterdam and were on their way home, (2) they were presumably between 16 and 70 years old (3) and they understood English. All nationalities were accepted and when people travelled together within a group, every member of that group was invited to participate in the survey and fill in the questionnaire individually because everyone has different experiences, emotions and sharing behavior.
Sample size -‐ In total 121 respondents filled in the questionnaire. A frequencies check was conducted to examine if there were any errors in the data. There were no errors found. Then a check for missing values was conducted and after that, 6 cases were deleted because of a large amount of missing or invalid data. The new data set contained 115 cases.
Demographics -‐ The respondents were mainly female (58.9%). The youngest participant was 16 years old and the oldest participant was 64 years old. The mean average age was 27.9 years and the median age of the sample was 26.0 years.
Nationalities -‐ The nationalities of the respondents are found in table 2. The vast majority (87.0%) lived in Europe. Because the other continents are not represented sufficient enough, the analysis was conducted without nationality considerations.
Continent Percentages
Africa 1.9%
America 6.5%
Asia 4.6%
Europe 87.0%
Table 2. Nationalities of the sample
Travel behavior -‐ The most stated purpose of the visit was holiday (93.9%), the other 6.1% had the main purpose of business. Both groups were included in this research. Most people stayed for four days in Amsterdam (45.5%), see table 3. 97.5% of the respondents stayed no longer than six days.
Number of nights Frequency Percentages
1 6 5.2% 2 2 1.7% 3 19 16.5% 4 50 43.5% 5 27 23.5% 6 8 9.1% 8 2 1.7% 12 1 0.9%
Mean average length of stay 4.2 Table 3. Length of stay of the respondents
Overall satisfaction -‐ The average level of overall satisfaction was 8.96, CI = [8.838, 9.082] (bootstrap) on a scale of 1 till 10.
Peak moments -‐ Every respondent had to fill in two peak moments on the questionnaire. For most part of the data analysis, a second data file was created in which the two peak moments were placed underneath each other in order to create a bigger data set and to evaluate the peak moments individually. This second data file contained 230 peak moments.
A frequency test was run for all variables and ten cases were deleted, due to missing data. 220 peak moments remained.
3.3 Pre-‐test
In order to test if there were any ambiguities or faults in the questionnaire, a pre-‐test was conducted among a group of twenty tourists that travelled by train to the airport. The pre-‐ test resulted in several eliminations and additions of questions and small textual modifications. A change was made for the word ‘leisure’ into ‘holiday’ and ‘stay in Amsterdam’ was changed to ‘stay in Amsterdam or areas nearby Amsterdam’. The question that asked people how likely it was that they would recommend visiting Amsterdam to their friends, was deleted since more than 90% of the respondents answered that question with “very likely” (very likely = 5 on a scale of 1 to 5). Also the preliminary results of pre-‐test appeared to be limited and therefore the questionnaire was extended with questions about sharing motivations and timing. However, the en route setting in the train seemed an appropriate research design.
3.4 Measurements
The questionnaire was structured around the following themes: demographics, two peak moments, experienced emotions, the use of social media, sharing motivations and the level of overall satisfaction. The next paragraphs describe which variables were tested and how these variables were constructed. An example of the final questionnaire can be found in appendix A.
Peak moments -‐ To test whether the Peak-‐end theory exists in a holiday setting, respondents were asked about their two peak moments of their holiday experience in Amsterdam (Question 3a & 3b). A person can only tell after an episode has finished, which moments cause the actual peaks (Fredrickson, 2000). Geng, et al. (2013) noted that the Peak-‐end theory proves to be a good prediction mechanism when tested on a short retention interval. Therefore, the survey was conducted immediately after the holiday experience has.
The Peak-‐end theory states that people evaluate affective experiences based on only two important moments that come to mind easily. According to Fredrickson, peak and end
moments ‘carry more personal meaning, such as emotions, than other moments’ (p. 589 in Fredrickson, 2000) and therefore are most likely to be remembered. In order to trace the peak and end moments, respondents were asked to describe their two ‘most memorable moments’ (Kemp, Burt & Furneaux, 2008). This definition was chosen because the words are understandable for most people and the definition is neutral in valence, which implicates that respondents could describe both positive and negative moments. To investigate whether one of the two peak moments took place during the end phase of their holiday, respondents were asked when the peak moments took place (Question 3b & 4b).
Emotions -‐ In order to investigate the effect of experienced emotions during peak moments on potential sharing behavior, respondents were asked to what extent they experienced eight different emotions during their two peak moments (table 4). Based on a selection of nine positive emotions (happy, grateful, amused, satisfied, proud, impressed, loving, interested and hopeful), derived from a study of Nawijn, et al. (2013), the pre-‐test tested which emotions respondents could possibly experience during peak moment. In the pre-‐test respondents also reported peak moments that contain negative emotions. Therefor the final questionnaire was adjusted and contained a broad panel of eight emotions (excited, relaxed, amazed, satisfied, sad, terrified, angry and unhappy). The selection of emotions was based on the eight affective descriptors ranging from pleasant to unpleasant, and arousing to non-‐arousing (Russell & Pratt, 1980, figure 2). Respondents were asked to rate all eight emotions on a 4-‐point scale (1 = not, 4 = strong), which was used by Richins (1997) (Question 3c and 4c). Figure 2. Circular ordering of affect descriptors (Russell & Pratt, 1980)
A principal component analysis was conducted with the eight emotions (excited, relaxed, amazed, satisfied, sad, terrified, angry and unhappy) in order to determine the underlying dimensions of positive and negative emotions, and high and low arousal emotions. An orthogonal rotation (varimax) was used (table 5). The results showed that three factors could be distinguished, as they had an eigenvalue over 1 and together explain 67.12% of the variance. The rotated component also showed that three factors could be created. Factor 1 was measured by two items (excited and amazed) and reflects the extent to which the respondents felt positive, high arousal emotions. Factor 2 was also measured by two items (satisfied and relaxed) and reflects the extent to which the respondents felt positive, low arousal emotions. Factor 3 was measured by four items (angry, sad, terrified and afraid) and reflects the extent to which the respondents experienced negative emotions.
A reliability analysis was done for all three factors. The Cronbach’s alpha for all the three factors was 0.7 or above (positive, high arousal emotions α = 0.703; positive, low arousal emotions α = 0.732, negative emotions α = 0.700), which implicated that the scales were reliable in measuring the constructs. The Cronbach’s alpha of factor 3 did not improve when one of the four items was deleted. The index of the three factors was constructed by the mean of the corresponding items.
Construct Categories Indicators
Positive emotions Positive, high arousal emotions Excitement
Arousal
Positive, low arousal emotions Satisfaction
Relaxed
Negative emotions
Negative, high arousal emotions
Terrified
Angry
Negative, low arousal
emotions
Sad
Unhappy
Table 4. Emotions constructs
Construct and item wording SL Factor 1: Positive, high arousal emotions (Cronbach’s alpha = 0.703)
Excited 0.81
Amazed 0.90
Factor 2: Positive, low arousal emotions (Cronbach’s alpha = 0.732)
Satisfied 0.82
Relaxed 0.90
Factor 3: Negative emotions (Cronbach’s alpha = 0.696)
Angry 0.73
Sad 0.63
Unhappy 0.76
Terrified 0.77
Table 5. Rotated component Matrix emotions
Social media -‐ Respondents were asked about their behavior or intention to share visual content on four different social media platforms: Facebook, Instagram, Snapchat and Whatsapp (question 3e and 4e). The pre-‐test showed that social media (Facebook and Instagram) and instant messaging (Whatsapp and Snapchat) were both used to share peak moments by more than 50% of the respondents. None of the respondents indicated that they used email for sharing visual content of peak moments. Because this research attempted to test sharing behavior of visual content, a selection was made of the four most popular social media platforms of 2016, which allow users to share visual content. Statista (April, 2016) showed that Facebook had 1,590 million active users, followed by Whatsapp with 1,000 million users. Also, Instagram with 400 million and Snapchat with 200 million were very popular social media platforms, exclusively driven by visual content. The platforms could be subdivided into two groups. Whatsapp and Snapchat belong to one group for which the audience reach can be limited and controlled by the sender, and allowing users to have a more private dialog (Munar & Jacobsen, 2014). The other group exists of Facebook and Instagram. Because these social media platforms usually have a
greater reach and provide the sender fewer options to control the audience, these platforms are more used to send a message, instead of having a conversation.
To gain more insight in sharing behavior of visual content on social media platforms, the survey also included a question about when visual content was shared on social media. Respondents could choose between four options (immediately, within 3 hours, within 12 hours or after more than 12 hours, Question 3g and 4g).
Sharing motivations -‐ Munar and Jacobsen (2014) tested six different motivations for sharing holiday experiences. The motivations were subdivided into self-‐related and community-‐related sharing motivations. Five of the six sharing motivations, proposed by Munar and Jacobsen (2014), were copied in this research (Question 3f and 4f). The main difference between this research and the study of Munar and Jacobsen is the focus on sharing peak moments, instead of sharing holiday experiences in general. Also, Munar and Jacobsen used a nominal measuring scale (yes, no and neither/nor), which resulted in nearly 50% of the respondents predominantly opting for the middle category ‘neither/nor’. Therefore a 4-‐point scale is used in this research (4 = disagree, little disagree, little agree and 1 = agree).
A principal component analysis (table 6) was conducted with the five items that measured sharing motivations. An orthogonal rotation (varimax) was used. The rotated components showed that two factors could be distinguished, as they had an eigenvalue above 1 and in combination explain 65.9% of the variance. Factor 1 consisted of the motivations: sharing impressions, encourage others and maintain social connections. The other two motivations, ‘being recognized for experiences’ and ‘informing others’, were included in factor 2. However, the Cronbach’s alpha for both factors appeared not to be sufficient enough (factor 1 α = 0.693 and factor 2 α = 0.463) to create a new reliable scale out of the items. Therefore, the items were continued to be used individually.
Construct and item wording SL
Factor 1
Share impressions 0.870 Encourage others to visit
Amsterdam
0.782
Maintain social connections 0.554
Factor 2
Being recognized for experiences 0.907
Inform others 0.556
Table 6. Rotated component Matrix sharing motivations
Overall satisfaction -‐ Overall satisfaction of tourists was measured on a 10-‐point scale (very dissatisfied = 1 to very satisfied = 10) (Dolnicar, Coltman & Sharma, 2013) in order to capture sufficient variance in the explanation of overall satisfaction.
Control variables -‐ In testing the hypotheses, several control variables were included in the survey: length of stay (measured by Question 1), purpose of travel (Question 2), demographics (age and gender) were measured in Question 6 and 8 and nationality (Question 7). Most of these questions were copied from the questionnaire of the Amsterdam Visitor Survey 2012. Education and income level were not asked, because the pre-‐test showed that people had difficulties answering these questions in public spaces.
3.5 Study design
Partly commissioned by Amsterdam Marketing, this research was carried out in Amsterdam, the main capital and most visited city of the Netherlands. By studying tourists of one specific destination and by collection data over a period of two weeks and varying the days, bias for this study was reduced (Munar & Jacobsen, 2014; Ryan & Glendon, 1998). The questionnaire was administered personally to the respondents, in order to gain higher response rates, faster results and a reduced chance of under-‐reporting. Geng, Chen, Lam and Zheng (2013) point to the fact that the Peak-‐end theory is a good explanation on short retention interval instead of long retention interval. Therefor the self-‐completion
questionnaire was conducted during the return journey, directly after the holiday to limit memory distortion probability (Kemp, Burt & Furneaux, 2008; Geng, Chen, Lam & Zheng, 2013). In order to be certain that respondents had finished their holiday experience in Amsterdam, the questionnaire was administered personally to the respondents in the train and respondents were asked if they had just visited Amsterdam and were on their way home. According to Rideng and Christensen (2004), the train is an appropriate spot to conduct surveys because it is a place with limited potential disruptions and no time pressure. Rideng and Christensen (2004) also found that train surveys have a relatively high response rate between 40% and 70%. The response rate for this research seemed even higher, presumably around 80%. There is no information about the non-‐respondents. The questionnaire was performed in English and contained a limited amount of questions to make participation more attractive and to retain the participant’s attention.
3.6 Data collection
Respondents were personally recruited at the train platform or in the train that ran from Amsterdam Central station to Schiphol Airport station and people who carried luggage were approached. The questionnaires were distributed in the train among tourists and picked up within five to ten minutes. There was no incentive.
3.7 Data analysis
The data of the questionnaire was processed through SPSS in order to determine the direction and significance of the relationships. A bootstrap with 500 samples was applied to test the stability of the estimations.
