696980 research-article2017 JTRXXX10.1177/0047287517696980Journal of Travel ResearchGuttentag et al. Empirical Research Articles Journal of Travel Research Why Tourists Choose Airbnb: A 1 –18 © The Author(s) 2017 Reprints and permissions: Motivation-Based Segmentation Study sagepub.com/journalsPermissions.nav hDttOpsI::/ /1d0o.i.1o1rg7/71/00.10147772/080745712877659167968960980 journals.sagepub.com/home/jtr Daniel Guttentag1, Stephen Smith2, Luke Potwarka3, and Mark Havitz3 Abstract Airbnb has grown very rapidly over the past several years, with millions of tourists having used the service. The purpose of this study was to investigate tourists’ motivations for using Airbnb and to segment them accordingly. The study involved an online survey completed in 2015 by more than 800 tourists who had stayed in Airbnb accommodation during the previous 12 months. Aggregate results indicated that respondents were most strongly attracted to Airbnb by its practical attributes, and somewhat less so by its experiential attributes. An exploratory factor analysis identified five motivating factors—Interaction, Home Benefits, Novelty, Sharing Economy Ethos, and Local Authenticity. A subsequent cluster analysis divided the respondents into five segments—Money Savers, Home Seekers, Collaborative Consumers, Pragmatic Novelty Seekers, and Interactive Novelty Seekers. Profiling of the segments revealed numerous distinctive characteristics. Various practical and conceptual implications of the findings are discussed. Keywords Airbnb, sharing economy, disruptive innovation, diffusion of innovations, segmentation Introduction assets (e.g., ride-hailing services like UberX) (Belk 2014; Botsman and Rogers 2010). The recent global economic Every night, hundreds of thousands of tourists choose not to recession helped catalyze the sharing economy, but it is also stay in a traditional tourism accommodation, such as a hotel, rooted in values related to sustainable consumption and com- but rather stay in the residence of a stranger found online via munity connectedness (Botsman and Rogers 2010; Chase Airbnb. The basic phenomenon of locals informally renting 2015; Stephany 2015). lodging to tourists has existed for centuries, but new Internet The rise of Airbnb and other peer-to-peer short-term and mobile technologies have revolutionized this practice rental services within the sharing economy represents a and allowed it to scale dramatically by facilitating virtual transformative innovation within the tourism accommoda- markets where communication and trust are established tion industry. By the summer of 2016 more than 100 million between hosts and guests (Guttentag 2015). Airbnb accom- guests had used Airbnb (Chafkin and Newcomer 2016), and modations typically involve an entire home (e.g., apartment, the service boasted over two million global listings (Airbnb house), or a private room in a residence where the host is also 2016). While it remains a topic of considerable debate, tradi- present. The Airbnb website (www.airbnb.com) is quite tional accommodations increasingly are viewing Airbnb as a straightforward: a prospective guest searches based on desti- significant threat (e.g., Martin 2016). nation, travel dates, and party size; the website returns a list of available spaces that can be refined by attributes like price, neighbourhood, and amenities; and then individual listings can be selected for greater detail, including a description, 1Hospitality and Tourism Management, Ryerson University, Toronto, photographs, and reviews from previous guests. Ontario, Canada Airbnb and other peer-to-peer short-term rental compa- 2Hospitality, Food, and Tourism Management, University of Guelph, nies (e.g., HomeAway, Wimdu) represent part of the broader Guelph, Ontario, Canada 3Recreation and Leisure Studies, University of Waterloo, Waterloo, “sharing economy” (also sometimes called “collaborative Ontario, Canada consumption”). The sharing economy is often associated with Internet and mobile technologies, and it involves con- Corresponding Author: Daniel Guttentag, Hospitality and Tourism Management, Ryerson sumers maintaining access to goods and services without University, TRS 3-062, 575 Bay Street, Toronto, Ontario M5G 2C5, owning them (e.g., bike-sharing), and ordinary individuals Canada. renting out or otherwise offering access to their underused Email: [email protected] 2 Journal of Travel Research 0(0) Because Airbnb is quite new, very limited research has an inner force or drive to satisfy an internal need (e.g., Gnoth investigated the important question of why tourists use it. 1999; Hawkins, Mothersbaugh, and Best 2007). As Dann Moreover, the existing research has portrayed Airbnb guests (1981) even acknowledged, “[Push motivation] deals with as a homogeneous group, thereby overlooking the likelihood tourist motivation per se” (p. 190). In contrast, pull motiva- that Airbnb users can be divided into market segments based tions are more aligned with the idea that consumers choose on their reasons for choosing the service. In fact, Airbnb list- products to seek certain benefits, and such benefits also serve ings are quite varied and the potential appeals of Airbnb as a common basis for customer segmentation (Haley 1968). include both practical advantages and experiential facets that There is little consensus within the tourism literature regard- may not generally go hand-in-hand, so the Airbnb market is ing the distinction between (pull) motivations and benefits. particularly suited for segmentation. Consequently, the pur- The present study is positioned as “motivation-based” pose of this study was to investigate tourists’ motivations for because the motivation terminology is somewhat more com- using Airbnb accommodations, and to segment them accord- mon within tourism literature, and some of the motivation ingly. A better understanding of guests’ motivations for using items considered were related to push factors. Airbnb, and of the segments identified and profiled, can offer Tourism studies segmenting on the basis of motivations valuable marketing insights for Airbnb, its hosts, and com- (or benefits) have often focused on visitation of a particular peting accommodation firms. Only with a clear understand- destination, attraction, or event. Motivation (or benefit)- ing of consumers’ reasons for choosing Airbnb can these based segmentation studies on accommodation choice are various entities make informed decisions regarding how best much more limited. Chung et al. (2004) used benefits sought to market toward Airbnb’s users, and even whether or not to segment independent business travelers staying in luxury such marketing efforts are worthwhile. The findings also are Seoul hotels, Inbakaran and Jackson (2005) used motivations useful for destination marketing organizations and other and some other variables to segment visitors to Australian tourism firms, as Airbnb guests’ motivations for using the hotel resorts, and Khoo-Lattimore and Prayag (2015) seg- service (e.g., seeking local authenticity) may highlight more mented “girlfriend getaway” travelers based on their prefer- general characteristics of their consumer preferences. ences for different accommodation attributes. Motivations to Use Airbnb Literature Review Tourists’ motivations for choosing Airbnb have been investi- Motivation-Based Market Segmentation gated by a handful of studies: Lamb’s (2011) unpublished Segmentation is the process by which a market is divided master’s thesis based on interviews with CouchSurfing and into groups that are internally similar in a meaningful way. Airbnb hosts, Guttentag’s (2015) conceptual examination of Segmentation serves as an important strategic tool for tour- Airbnb through the lens of disruptive innovation, Tussyadiah’s ism marketers, providing them with actionable insights on (2015) and Tussyadiah and Pesonen’s (2016) surveys of peer- targeting, positioning, and competitive analysis (Dolinicar to-peer short-term rental users from the perspective of col- 2012). In post hoc segmentation, quantitative data analysis laborative consumption, Quinby and Gasdia’s (2014) survey techniques generate a classification system based upon a col- of peer-to-peer short-term rental users (for the tourism lection of variables, often focusing on consumers’ purchase research company Phocuswright; see Hennessey 2014), and motivations (Dolnicar 2002). Nowak et al.’s (2015) survey of Airbnb users (for the finan- The term motivation has been defined in different ways, cial services company Morgan Stanley). These studies have but essentially refers to the reasons why someone engages in identified a range of potential motivations. Price (or economic a particular behavior (Hawkins, Mothersbaugh, and Best benefits) has been recognized by all of the studies listed 2007). Tourism literature generally has adopted Dann’s above, sometimes as the most important motivator (Nowak (1977, 1981) push–pull motivation framework that recog- et al. 2015; Tussyadiah 2015) but other times as somewhat nizes both the internal drives that inspire someone to travel less important (Lamb 2011; Quinby and Gasdia 2014). (“push factors”) and the particular characteristics of a certain Household amenities and space have additionally been travel product that persuade the traveler to choose it (“pull acknowledged in several studies and actually were the two factors”). While conceptually distinct, push and pull factors top motivations found by Quinby and Gasdia (2014). are often closely related (Kim, Noh, and Jogaratnam 2007). Authenticity also has been highlighted by several studies, This study focused on tourists’ choice of Airbnb as pertain- including by Lamb (2011), who posed it as the primary driver ing to particular characteristics of Airbnb accommodations, of Airbnb use, and by Nowak et al. (2015), who found it to be and was therefore focused on pull motivations, although sev- one of the strongest motivations. Also, Guttentag (2015) eral of the motivation items were related to push factors. viewed interacting with locals as a part of authenticity, but Push motivations are aligned with a more precise concep- Tussyadiah (2015) and Tussyadiah and Pesonen (2016) tualization of motivation (or motive), common within psy- positioned such interaction separately, as part of a social chology, consumer behavior, and some tourism literature, as benefit enjoyed from using Airbnb. Tussyadiah (2015) also Guttentag et al. 3 highlighted the importance of sustainability as a motivation to home swaps (e.g., Andriotis and Agiomirgianakis 2014), and use Airbnb. Finally, Nowak et al. (2015) considered location, CouchSurfing (e.g., Bialski 2011). which they actually found to be the second most important motivation. Innovation Adoption These studies provide some valuable insights into why tourists choose Airbnb, but this body of research also suffers The tourism accommodation choice literature highlights var- from numerous limitations. It is a fairly small body of litera- ious motivations that may draw users to Airbnb. However, ture with minimal peer-reviewed empirical research, and the that literature has been largely atheoretical, so two innova- studies have reached somewhat incongruent conclusions. tion concepts—disruptive innovation and the diffusion of Additionally, the studies have tended to be limited in the innovations—were used to add a conceptual foundation to breadth of possible motivations they considered. Furthermore, the present study. These concepts were drawn on for addi- several of the studies examined peer-to-peer short-term rent- tional guidance on variables to consider, and to better under- als in general, instead of a particular company like Airbnb, stand these different variables. which may have obfuscated findings because of dissimilari- As described by Christensen (1997) and Christensen and ties between different peer-to-peer short-term rental services. Raynor (2003), a disruptive innovation is a product whose Finally, all of the studies viewed Airbnb users as homoge- appeal does not derive from improved performance, which nous, rather than as members of potential motivation-based one may normally expect, as disruptive innovations rather market segments. The present research, therefore, provides a underperform in comparison with prevailing products’ key valuable contribution to this area of inquiry by considering a attribute(s). Nevertheless, disruptive innovations introduce broad range of motivations, by focusing specifically on an alternative package of benefits, generally centered on Airbnb, and by recognizing the potential for different moti- being cheaper, simpler, smaller, and/or more convenient. In vation-based segments of Airbnb users. other words, disruptive innovations are inferior “good enough” products when compared directly to existing com- petitors, but their unique set of attributes modifies the pre- Tourism Accommodation Choice vailing value proposition in a way that appeals to some Although very few studies have focused on Airbnb choice, consumers. The notion of disruptive innovation seems to myriad researchers have investigated tourism accommoda- apply well to Airbnb accommodations—they will seemingly tion choice more broadly. Most of this research has exam- underperform traditional accommodations when considering ined hotel choice, frequently with respondents rating the conventional attributes like cleanliness and security, but they importance of different hotel attributes (e.g., Lockyer tend to be relatively inexpensive, can foster a more authentic 2005; Sohrabi et al. 2012). This literature has identified a local experience, and can provide various benefits associated variety of key attributes influencing hotel decisions, with staying in a home (e.g., household amenities) (Guttentag including cleanliness, location, reputation, price, value, 2015). In other words, Airbnb offers a new value proposition service quality (e.g., staff friendliness and helpfulness), that will appeal to some consumers. room comfort, and security (Chu and Choi 2000; Dolnicar This notion that disruptive products introduce an alterna- and Otter 2003). tive package of benefits offers a basic explanation of the con- Complementing the hotel choice literature is a more limited sumer demand for such products. It is essentially a literature on the choice to use non-hotel forms of accommoda- Lancastrian approach of decomposing products into collec- tion (e.g., bed-and-breakfasts, homestays). Whereas the hotel tions of attributes (Lancaster 1966). Nevertheless, the most choice literature has focused on the choice between hotel concentrated look at disruptive innovation demand comes properties, the non–hotel choice literature has focused on the from Adner (2002), who modeled demand for computer disk choice to use these alternative forms of accommodation more drives and demonstrated the particular importance of unit generally. This literature has tended to highlight the unique price. Adner noted that as product performance levels nature of the experience, rather than merely the practical attri- become very high, market heterogeneity is reduced because butes that dominate the hotel choice literature. For example, most consumers are satisfied with performance, and the McIntosh and Siggs (2005) found that alternative accommo- characteristics that previously distinguished them become dation guests enjoyed the unique character and homely feel of decreasingly relevant. In turn, unit price, rather than a price– the accommodations, the personalized service and personal performance ratio, becomes increasingly important in interaction with the hosts, and the opportunity to receive use- encouraging adoption. ful local knowledge from the hosts. Likewise, Stringer (1981) Additional insight into innovation adoption can be found researched guests of British bed-and-breakfasts and found in literature on the diffusion of innovations, which broadly they were drawn by both the experience and the economical examines the spread of innovations as they are increasingly price. Similar findings highlighting the importance of interper- adopted by members of a society. This literature has high- sonal and authentic experiences, in addition to saving money, lighted the significant influence certain innovation attributes have been found in research on homestays (e.g., Wang 2007), have over adoption decisions. Of particular importance is 4 Journal of Travel Research 0(0) “relative advantage,” which refers to the perception that an various characteristics of a “hard-to-reach” population innovation is better than its predecessor (Arts, Frambach, (Marpsat and Razafindratsima 2010). A multiple-frame and Bijmolt 2011; Evanschitzky et al. 2012; Rogers 2003). online non-random sampling approach therefore was Such benefits can vary widely and include financial implica- deemed necessary. The majority of the respondents were tions, functional attributes, social prestige, convenience, sat- recruited via six travel-related Facebook groups based isfaction, or immediacy of reward (Rogers 2003). Whereas around major Canadian cities, and consisting of thousands disruptive innovation tends to focus on objective functional of members apiece. Additionally, respondents were recruited performance, the broader perspective offered by relative via Mechanical Turk (MTurk), an opt-in online panel that is advantage highlights important indirect advantages of prod- increasingly being used in social science research. As rec- uct adoption. For example, the notion of prestige is reminis- ommended by Chen (2012) and Kittur, Chi, and Suh (2008), cent of tourism “bragging rights,” which Kerr, Lewis, and data quality from the MTurk responses was promoted by Burgess (2012) suggested influence some travelers’ destina- paying a relatively high compensation (these respondents tion choice. In addition to relative advantage, innovations are were paid per completion, rather than entered in the lottery more appealing if they are “compatible” with an adopter’s draws), including two verifiable questions, and restricting values, beliefs, positive past experiences, and existing needs respondents to certain countries (the United States, the (Arts, Frambach, and Bijmolt 2011; Rogers 2003; Tornatzky United Kingdom, Australia, and New Zealand). A handful of and Klein 1982). other sampling approaches additionally were used to further Beyond characteristics of the innovation itself, innovation bolster and diversify the sample. These approaches involved adoption decisions also are influenced by characteristics of the publishing invitation messages on travel-themed Facebook potential adopter. “Innovativeness” refers to how early an pages, travel-themed Twitter feeds, and an Airbnb-focused individual tends to be in adopting innovations. Innovativeness page on the website Reddit; sending invitation messages to is sometimes examined using chronological adopter segments a small number of Airbnb hosts and asking them to forward (“early adopters,” “laggards,” etc.) (Rogers 2003), and some- the invitation to their recent guests; sending invitation mes- times as a continuum-based personality trait (Midgley and sages to travel bloggers who had recently used Airbnb; and Dowling 1978). Although innovativeness has not received including a referral link at the end of the survey. widespread attention from tourism scholars, a handful of stud- Although the sampling approach was non-random, the ies have found different forms of innovativeness were posi- combination of different sampling frames was intended to tively associated with various purchase behaviors (Couture reduce the overall study sample bias. Also, both Facebook et al. 2015; Lee, Qu, and Kim 2007; San Martín and Herrero and MTurk, from which the majority of the sample was 2012). Innovativeness is very closely related to the notion of drawn, have been recognized as recommendable sampling novelty-seeking (Hirschman 1980), which is a concept more frames that produce high-quality data on a level that is gener- common within the tourism literature. Conceptualized as a ally comparable to or better than many common alternatives desire for new and unfamiliar stimuli (Lee and Crompton (Baltar and Brunet 2012; Buhrmester, Kwang, and Gosling 1992; Snepenger 1987), novelty-seeking has been central to 2011; Ramo and Prochaska 2012; Simons and Chabris 2012). some classic tourism typologies (Cohen 1972; Plog 1974) and Moreover, as compared to the general population, many of has been used in various tourism segmentation studies (e.g., the biases characterizing users of websites like Facebook, Chang, Wall, and Chu 2006; Mo, Havitz, and Howard 1994). MTurk, and Reddit should be consistent with biases found among users of an online service like Airbnb. Methods Survey Design Data Collection The survey items were primarily Likert scale and multiple- Individuals who had used Airbnb during the previous 12 choice. The questions focused chiefly on a respondent’s most months were recruited to complete an online survey, with recent Airbnb stay in order to minimize confusion. A pretest data collection beginning in July 2015 and concluding in was conducted with several prior Airbnb guests who were October 2015. Two Amazon gift cards of US $50 apiece (or members of the principal researcher’s social circle, and it its international equivalent) were offered as incentives, and involved completing the survey and answering a series of were distributed in lottery draws. Respondents needed to open-ended questions regarding possible issues like confu- have been significantly involved in the decision to choose sion and fatigue. Questions regarding Airbnb use, trip char- Airbnb accommodation, and only one travel party member acteristics, and sociodemographics were asked in a (from a respondent’s most recent Airbnb stay) could com- straightforward manner. However, household income level plete the survey. was asked using a Likert scale –“Well below average” to Because Airbnb is relatively new, has only been used by “Well above average” relative to a respondent’s home coun- a small proportion of the population, and has not been try – as this approach accommodated respondents from dif- widely researched, the desired respondents exhibited ferent countries earning income in different currencies. Guttentag et al. 5 Agreement with different potential motivations for choos- respondents from the Canadian Facebook groups, MTurk, ing Airbnb was measured using a 17-item Likert scale (1 = and all other sampling frames were compared along a series strongly disagree to 6 = strongly agree). As an exploratory of variables using one-way analysis of variance (ANOVA), study, the items were written uniquely for this research. They Welch, and chi-square tests. were based primarily on the previously described motiva- An exploratory factor analysis, using principal axis fac- tions that have been proposed in prior motivation research on toring extraction and a direct oblimin oblique rotation, was Airbnb and the broader peer-to-peer short-term rental sector. then run on the 17 Airbnb motivations to identify underlying In addition, some guidance was derived from the accommo- factors, with the goal of easing interpretation of subsequent dation choice literature, especially studies looking at the analyses of the motivation data. Although tourism research- choice to use alternative forms of accommodation. Finally, ers often simply extract factors with eigenvalues above one, the concepts of disruptive innovation and the diffusion of this approach is problematic (Ledesma and Valero-Mora innovations, and relevant studies on these topics, were relied 2007), and was inappropriate for the present study because on for additional direction when designing the scale. The 17 of the size of the communalities after extraction (Field 2013). motivation items used pertained to six dimensions—price, Rather, guidance on the number of factors to extract was functional attributes, unique and local authenticity, novelty, based on an examination of the scree plot and a parallel anal- travel bragging, and sharing economy ethos. ysis performed using the psych package in R (Revelle 2015). A price item was framed in terms of Airbnb’s compara- Subsequently, a cluster analysis involving the 17 motiva- tively low cost relative to other accommodation options, as tion items was undertaken. Prior to conducting the cluster this straightforward comparative price attribute is central to analysis, multicollinearity was assessed by verifying that no the notion of disruptive innovation (Adner 2002; Christensen clustering variables exhibited correlations above 0.9 (Sarstedt 1997). Five items relating to functional attributes were and Mooi 2014). The cluster analysis employed the two- included (e.g., location convenience, household amenities) stage cluster approach (Punj and Stewart 1983) that has been and were based on the existing Airbnb and non-hotel accom- used widely by tourism researchers (e.g., Chang 2006; modation choice literature (e.g., Guttentag 2015; McIntosh Prayag and Hosany 2014). The two-stage cluster approach and Siggs 2005). Four items were included regarding the involves initially conducting a hierarchical cluster analysis desire for unique and authentic local experiences (e.g., inter- and subsequently entering some of the resulting parameters action with host/locals, staying in a non-touristy area). These into a k-means analysis. Ward’s method with squared items were again based on the existing Airbnb and non-hotel Euclidean distance was used for the agglomerative hierarchi- accommodation choice literature (e.g., Bialski 2011; Lamb cal clustering. The percentage change in heterogeneity within 2011). Three items associated with novelty-seeking were clusters in subsequent clustering stages, as indicated by the included, based on Lee and Crompton’s (1992) research on agglomeration coefficient, was initially examined for guid- novelty-seeking in tourism. Those authors identified four ance on the optimal number of clusters to specify for the novelty-seeking dimensions—thrill, change from routine, k-means analysis (Hair et al. 2014). Following the k-means boredom alleviation, and surprise—and one item associated analysis, the variance ratio criterion (Sarstedt and Mooi with three of these dimensions was included, with boredom 2014) and hit ratios from discriminant analyses were used for alleviation excluded because it applies more to travel (push) guidance on the final number of clusters. motivations than accommodation choice (pull) motivations. A variety of profiling variables then were used to compare Three items related to the ethos of the sharing economy were the segments. Chi-square, one-way ANOVA, and Welch tests included (e.g., Airbnb’s environmental friendliness), and were conducted to assess differences between the segments. were based on the general sharing economy literature (e.g., In cases of statistical significance, standardized residuals Botsman and Rogers 2010) and Tussyadiah’s (2015) peer-to- (chi-square), Gabriel’s tests (one-way ANOVA), and Games- peer short-term rental study. Finally, one item on travel brag- Howell tests (Welch) were used to better identify segment ging was included, centered on tourists’ potential desire to differences. The variables Number of Nights, Number of have an experience they could tell friends and family about. Other Guests, and Total Times Used Airbnb were logarithmi- This item was based on prior use of travel bragging in seg- cally transformed prior to the analyses in order to account for mentation studies by Cha, McCleary, and Uysal (1995) and a high positive skew. Also, to limit the influence of extreme Sirakaya, Uysal, and Yoshioka (2003). values, six durations that exceeded 30 nights were changed to 31 prior to the analysis (Field 2013). Data Analysis Results Various analyses were employed to answer the research questions guiding this study. All analyses were conducted Sample Profile using SPSS, Excel, and R software. To begin, basic descrip- tive statistics were used to obtain a general overview of the A total of 923 surveys were received. Data screening elimi- sample and the responses to the different survey items. Also, nated numerous surveys as a result of issues such as 6 Journal of Travel Research 0(0) incompleteness, carelessness (indicated by an especially Factor Analysis short time spent on the survey or a high number of consecu- An exploratory factor analysis was performed on the 17 tive identical responses) (Curran 2016), and incorrect Airbnb motivations. Initial examinations of the inter-item answers to the verifiable MTurk questions. The final sample correlation matrix led to the removal of two items (“low used for the analyses consisted of 844 total respondents. Of cost” and “location convenience”) for which all correlations these, 72.4% were derived from the Canadian travel-themed were much lower than the common threshold of 0.3 (Field Facebook groups, 16.4% were derived from MTurk, 10.3% 2013). Subsequently, an initial run of the exploratory factor were derived from other sampling frames (e.g., Reddit, analysis led to one item, unique (nonstandardized), cross- referrals), and 0.9% were of unspecified origin. When loading onto two factors with similar factor loadings, so this respondents from the Canadian Facebook groups, MTurk, item also was removed. and all other sampling frames were compared, significant The remaining 14 variables were shown to be appropriate differences were detected along some variables (e.g., gen- for factor analysis—Cronbach’s alpha was 0.868 (N = 814), der, trip purpose), whereas for others the groups were found the Kaiser–Meyer–Olkin measure of sampling adequacy had to be fairly similar (e.g., age, type of Airbnb accommoda- a very high value of 0.890, the Kaiser–Meyer–Olkin values tion used). for the individual items were all at least 0.736, and Bartlett’s Characteristics of the overall sample can be observed in test of sphericity was significant (χ2(91) = 4085.74, p < Table 1. As can be seen, 67.8% of the respondents were 0.001). Parallel analysis recommended both four- and five- female, 81.9% were between the ages of 21 and 40, 92.8% factor solutions, and the five-factor solution was chosen had at least a university or college degree, and 77.8% per- because it was more clearly suggested by the scree plot and ceived their household financial status as at least “just above because the four-factor solution combined two seemingly average” in their home country. Owing to the sampling conceptually distinct factors in a way that led to fairly low frames used, 74.3% of the respondents resided in Canada and factor loadings for one of the factor’s items. Moreover, 23.0% resided in the U.S. For their most recent Airbnb stay, because the goal of the factor analysis was to identify latent 80.3% had been traveling for leisure, 59.7% were on an structures among the motivations in order to ease interpreta- international trip, 18.1% perceived themselves as “back- tion of the subsequent cluster analysis, the creation of more packers,” 70.3% were staying in an entire home (rather than precise factors was preferable. The final five-factor solution sharing a residence with the host), 62.5% were staying for was very clean and explained 69.1% of the total variance. All between two and four nights, 75.5% were staying with factor loadings easily exceeded the commonly used criterion between one and three other accompanying guests, and of 0.32 (Tabachnick and Fidell 2013), except for one item 57.6% were staying with a spouse or partner. Finally, 55.8% with a loading of 0.26, yet even that value could still be con- had used Airbnb no more than three times, 57.7% had first sidered significant given the size of the sample (Stevens used Airbnb in 2014 or 2015 (data collection occurred 2009). between July and October 2015), and 9.9% had experience The factor analysis results can be observed in Table 2. The as Airbnb hosts. first factor, Interaction, explained a large share of the vari- Because this study used nonprobability sampling, to ance (38.4%), and consisted of two items associated with assess the general representativeness of the sample various interacting with one’s host or other locals. The second factor, sample characteristics were compared with those of Airbnb’s Home Benefits, explained 10.8% of the variance and con- guest population that could be gleaned from the roughly 25 sisted of three items associated with staying in a home. The local economic impact reports that Airbnb has published third factor, Novelty, explained 8.7% of the variance and (e.g., Airbnb 2015b), and a report on its guests during the consisted of the three novelty items based on Lee and summer of 2015 (Airbnb 2015a). Airbnb stated in its sum- Crompton’s (1992) work and the single travel bragging item. mer 2015 report that 54% of its guests were female (Airbnb The fourth factor, Sharing Economy Ethos, explained 6.0% 2015a), in comparison with 67.8% of the present study’s of the variance, and consisted of the same three items origi- respondents. In the same report, Airbnb claimed that its nally proposed for this construct. Finally, the fifth factor, average guest age was 35 (Airbnb 2015a), and if one esti- Local Authenticity, explained 5.3% of the variance, and con- mates the mean age of the present study’s respondents using sisted of two items associated with having an authentic local the midpoint of each age group (e.g., 35 for 31–40), the experience. result is an average age of 33. Airbnb economic impact reports suggest that about 86% of its visitors are traveling Cluster Analysis for leisure, in comparison with 80.3% for the present study. Airbnb economic impact reports and claims to the media Tourism segmentation research has frequently used a factor- (Lu 2015) both have suggested that guests’ average length cluster approach, in which variables are first reduced via fac- of stay is 4.5 nights, and the average length of stay for tor analysis and then the resulting factor scores are used for respondents in the present study was 4.54 nights. the cluster analysis. However, this procedure is discouraged Guttentag et al. 7 Table 1. Sample Characteristics. Characteristics % n Characteristics % n Gender Type of Airbnb accommodation used Female 67.8 553 Entire place 70.3 586 Male 32.1 262 Private bedroom 27.6 230 Transgender 0.1 1 Shared space 2.2 18 Age Nights ≤20 1.1 9 1 9.5 79 21–30 52.3 437 2 22.0 183 31–40 29.7 248 3 23.6 196 41–50 8.0 67 4 16.8 140 51–60 5.6 47 5 9.7 81 ≥61 3.3 28 6 5.4 45 Highest level of education 7 6.0 50 High school or less 7.2 59 8–29 5.7 47 University / college 62.6 510 ≥30 1.2 10 Graduate / professional 30.2 246 Number of other guests Household financial status (relative to home country) 0 11.2 93 Well below average 1.0 8 1 50.4 417 Below average 5.4 42 2 12.3 102 Just below average 15.8 123 3 12.8 106 Just above average 46.9 365 4 5.7 47 Above average 27.7 216 5 4.3 36 Well above average 3.2 25 6+ 3.3 27 Country of residence Type of other guests (% of total sample) Canada 74.3 589 Spouse/partner 57.6 486 USA 23.0 182 Child(ren) 10.9 92 Other 2.8 22 Friend(s) 31.0 262 Trip purpose Professional colleague(s) 2.0 17 Business 3.5 29 Total times used Airbnb Convention, conference, event 7.5 63 1 22.0 182 Leisure 80.3 673 2 16.7 138 Visiting friends/family 8.7 73 3 17.1 142 Destination region 4 10.9 90 Canada 23.0 194 5 8.9 74 Europe 28.9 244 6 5.9 49 USA 36.4 307 7 4.3 36 Other 11.6 98 8–10 7.5 62 Destination type 11+ 6.6 55 Domestic 40.3 319 Year first used Airbnb International 59.7 473 2008–2010 4.0 33 Self-described “backpacker” on trip 2011 6.6 55 No 81.9 685 2012 12.7 105 Yes 18.1 151 2013 19.0 158 2014 32.0 266 2015 25.7 213 Ever been an Airbnb host No 90.9 758 Yes 9.1 76 Note: “Business” signifies business (other than convention, conference, or other major event). “Shared space” refers to sleeping in a shared area, such as a futon in the host’s living room. because a large quantity of meaningful variance is lost in the Dolnicar et al. (2014), the present study’s sample size was factor analysis (Dolnicar and Grün 2008). Fortunately, as per large enough to include all 17 motivation items in the cluster recommended respondent-to-variable ratios stated by analysis. 8 Journal of Travel Research 0(0) Table 2. Factor Analysis of the Motivations to Choose Airbnb. Factor Factor Pct. Variance Explained Cronbach’s Average of the Motivation Loadings Eigenvalues (Cumulative) α Mean Scores Interaction 5.37 38.36 0.78 3.68 To interact with host, locals .79 (38.36) To receive useful local info/tips from my host .71 Home benefits 1.51 10.79 0.65 4.42 For the large amount of space .66 (49.15) For the access to household amenities .65 For the homely feel .47 Novelty 1.21 8.65 0.80 3.53 I thought the experience would be exciting .78 (57.80) To do something new and different .75 To have experience I could tell friends/family .64 about I thought the experience would be .55 unpredictable Sharing Economy Ethos .83 5.96 0.73 3.62 I wanted the money I spent to go to locals .87 (63.76) Staying with Airbnb is environmentally friendly .60 I prefer the philosophy of Airbnb .45 Local Authenticity .75 5.33 0.63 4.39 To have an authentic local experience .71 (69.09) To stay in a non-touristy neighborhood .26 Note: All items were measured on a six-point Likert scale ranging from 1 = strongly disagree to 6 = strongly agree. For the Interaction and Local Authenticity factors, the reported Cronbach’s α score is the “Cronbach’s α based on standardized items,” which is equivalent to the Spearman’s-Brown coefficient and is a more appropriate reliability measure for two-item scales (Eisinga, Grotenhuis, and Pelzer 2013). Initial clustering of the data resulted in cluster solutions Table 3 displays the group means for the selected five- that essentially grouped the respondents into segments of cluster solution, in addition to overall sample means. To strong, medium, and low levels of agreement across all of the ease interpretation, the motivations are organized in accor- motivations. Such results appeared chiefly reflective of dance with the factor analysis (Table 2). However, the three response-style effects associated with respondents’ different motivations that were excluded from the factor analysis baseline levels of agreement (Hair et al. 2014). Consequently, were reinserted into Table 3, as they were included in the each individual’s responses were standardized via within cluster analysis. The “low cost” and “location convenience” case z-scores, which is a recommended transformation in motivations were added to the top of the motivation list, such circumstances (Hair et al. 2014; Schaninger and Buss and the “unique (non-standardization)” motivation was 1986). Standardizing scores by case is fairly comparable to added to the Novelty factor upon which it loaded most using correlation as a distance measure in hierarchical clus- heavily. Also to ease interpretation, the cell values were tering (Hair et al. 2014; Sarstedt and Mooi 2014), and while shaded based on their deviations from the sample mean for it leads to some meaningful variance being lost, it can render each variable, with darker shades indicating higher levels a solution that is more interpretable, more heterogeneous, of agreement compared to the other segments. The F-values and more clearly related to external variables (Schaninger in Table 3 display the results of univariate ANOVAs com- and Buss 1986). paring the mean scores for each segment. These values Agglomeration coefficients were examined for guidance function primarily as indicators of the degree to which each on the optimal number of clusters (Hair et al. 2014), but there motivation contributed to the final cluster solution (SPSS was no clear cut-off point. Therefore, cluster centroids were 2016). The associated p-values have not been included saved for three, four, five, six, and seven-cluster solutions, because in k-means analysis the clusters are selected to and imported into a k-means analysis for further examina- maximize differences between clusters, so the p-values tion. Subsequently, based on the variance ratio criterion, hit should not be interpreted as tests of the hypothesis that the ratios from discriminant analyses, and an examination of cluster means are equal (SPSS 2016). (non-transformed) variable means for various cluster solu- Multiple discriminant analysis was used to help confirm tions, a five-cluster solution was selected. the validity of the cluster solution. The five clusters were e F 3.40 5.57 3.86 2.40 2.29 6.05 4.78 4.04 1.37 9.73 4.16 8.03 1.42 9.59 6.18 6.60 7.51 sampl 6 4 16 9 10 7 3 6 7 3 6 4 4 1 1 1 1 he m t aborative Pragmatic Interactive TotalnsumersNovelty SeekersNovelty Seekers (N = 807) 54 / 19.1%)(n = 175 / 21.7%)(n = 138 / 17.1%)M, SD 5.285.165.045.22, 0.95 4.885.034.934.99, 1.00 4.762.574.593.45, 1.51 4.883.034.783.90, 1.43 3.534.373.644.14, 1.39 4.185.044.204.71, 1.30 4.664.544.514.41, 1.30 4.174.784.894.06, 1.29 4.234.854.864.03, 1.38 3.474.154.083.41, 1.40 2.322.933.622.63, 1.27 4.814.914.884.35, 1.34 4.773.583.513.70, 1.36 3.933.362.973.24, 1.30 4.563.873.793.91, 1.34 5.084.564.964.45, 1.29 5.114.614.184.32, 1.41 The cluster mean scores were shaded according to their deviation froh, except for the two extreme intervals that extend indefinitely. CollMoney SaversHome SeekersCo (n = 188 / n = 152 / 18.8%)23.3%)(n = 1 5.675.01 5.174.91 2.443.19 3.093.92 3.645.20 4.365.52 3.135.05 2.883.66 2.863.46 2.432.96 2.172.23 2.774.35 2.863.76 2.573.30 3.114.17 3.234.47 3.264.39 m 1 = strongly disagree to 6 = strongly agree. shading intervals of 0.2 standard deviations eac ( nging froe are 10 Table 3. The Motivation-Based Cluster Solution. Factor Motivation For its comparatively low cost For the convenient location Interaction To interact with host, locals To receive useful local info/tips from my host Home benefits For the large amount of space For the access to household amenities For the homely feel Novelty I thought the experience would be exciting To do something new and different To have experience I could tell friends/family about I thought the experience would be unpredictable To have a unique (nonstandardized) experience Sharing Economy Ethos I wanted the money I spent to go to locals Staying with Airbnb is environmentally friendly I prefer the philosophy of Airbnb Local Authenticity To have an authentic local experience To stay in a non-touristy neighborhood Note: All items were measured on a six-point Likert scale ramean, with darker shading indicating higher agreement. Ther 9 10 Journal of Travel Research 0(0) Table 4. Summary of the Discriminant Analysis Results. Percent Cumulative Canonical After Wilks’ Function Eigenvalue Variance Percent Correlation Function Lambda Chi-square df p 1 1.403 38.2 38.2 0.764 0 0.081 1995.223 68 <0.001 2 1.105 30.1 68.3 0.725 1 0.195 1298.127 48 <0.001 3 0.854 23.2 91.5 0.679 2 0.411 706.292 30 <0.001 4 0.312 8.5 100.0 0.487 3 0.762 215.660 14 <0.001 Note: 92.8% of original cases correctly classified. used as the dependent variable, with the 17 motivation regards to the type of Airbnb accommodation used, the items acting as the predictor variables. Importantly, the dis- number of nights stayed in the Airbnb accommodation, the criminant analysis used the raw, nontransformed motiva- number of other guests present in the accommodation, and tion scores, rather than the z-scores used in the cluster whether or not they were accompanied by children. The seg- analysis. The discriminant analysis generated four discrim- ments did not differ significantly with regards to whether or inant functions, shown in Table 4. As can be observed, the not they were accompanied by spouses/partners or friends. four discriminant functions in combination significantly Finally, significant differences were found when looking at differentiated the groups, as did all other subsequent com- several variables related to Airbnb usage history (Table 6)— binations generated by peeling away the functions one at a the total number of times they had used Airbnb, the year time. Also, the discriminant analysis correctly classified they first used Airbnb, and whether or not they had ever 92.8% of the cases, which is a high hit ratio that lends sup- been an Airbnb host. port to the cluster solution. Before considering the different segments, it is worth- Money Savers while to describe the aggregate levels of agreement with the various motivations (Table 3). Respondents on average The Money Savers were chiefly attracted to Airbnb by its agreed with nearly all of the proposed motivations (3.5 comparatively low cost. They agreed more strongly with this was the mathematical midpoint of the six-point scale). By motivation than any other segment agreed with any other a fairly substantial degree, respondents agreed most motivation. The Money Savers exhibited a neutral opinion or strongly with the “low cost” motivation. That was fol- disagreement with most of the other motivations considered. lowed by the “location convenience” and “household ame- Money Savers tended to be somewhat young, with 62.9% nities” items. Agreement also was relatively high with the aged 30 and under (vs. a 53.2% average), and were signifi- other two Home Benefits items and the two Local cantly less likely than average to be traveling with children Authenticity items. Agreement with the Novelty items was (3.3%, vs. a 10.3% average). mixed, as respondents moderately agreed with several items and disagreed with two others. Agreement with the Home Seekers Sharing Economy Ethos and Interaction items was fairly limited. The Home Seekers were especially motivated by the three Home Benefits items. They agreed more with these items than with the “low cost” item, representing the only instances Cluster Profiles in which a segment agreed with anything more than low cost. Based on their motivations for choosing Airbnb, the five The Home Seekers were significantly older than average clusters were named Money Savers, Home Seekers, (23.7% aged 41 and older, vs. a 16.8% average), were the Collaborative Consumers, Pragmatic Novelty Seekers, and most well educated (35.4% held a graduate or professional Interactive Novelty Seekers (Table 3). A variety of profiling degree, vs. a 29.7% average), and were significantly less variables were used to better understand the different seg- likely than average to be backpackers (10.2%, vs. a 17.8% ments, the results of which can be seen in Tables 5 and 6. average). They also were significantly more likely than aver- With regards to demographics (Table 5), the segments dif- age to be renting an entire home (92.0%, vs. a 71.0% aver- fered significantly by age, but not by gender, highest level of age), were using Airbnb for significantly longer stays than all education completed, or household financial status. When other segments (5.72 nights, vs. a 4.24 average), had the looking at trip characteristics of the most recent Airbnb stay highest average number of accompanying guests (2.27, vs. a (Table 5), the segments differed with regards to backpacker 1.79 average), were the most likely to be staying with a status, but not trip purpose or destination type. When con- spouse/partner (64.9%, vs. a 57.6% average), and were sig- sidering accommodation usage characteristics of their most nificantly more likely than average to be staying with chil- recent Airbnb stay (Table 6), the segments differed with dren (22.3%, vs. a 10.3% average). They also had used