Market accessibility and hotel prices in the Caribbean: The moderating effect of quality-signaling factors

Tourism Management 56 (2016) 40e51

Contents lists available at ScienceDirect

Tourism Management journal homepage: www.elsevier.com/locate/tourman

Market accessibility and hotel prices in the Caribbean: The moderating effect of quality-signaling factors Yang Yang a, *, Noah J. Mueller b, Robertico R. Croes c a

School of Tourism and Hospitality Management, Temple University, Philadelphia, PA 19122, USA Department of Geography, University of Florida, Gainesville, FL 32611, USA c Rosen College of Hospitality Management, University of Central Florida, Orlando, FL 32819, USA b

h i g h l i g h t s
We investigated the influence of market accessibility on hotel prices.
We employed a three-level mixed-effect linear regression model.
The influence of accessibility is moderated by various quality-signal factors.
Increasing hotel's reputation helps buffer the impacts of being inaccessible.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 26 February 2014 Received in revised form 31 December 2015 Accepted 14 March 2016

The purpose of the paper is to investigate the influence of market accessibility on hotel prices and how this influence is moderated by various quality-signaling factors, such as online user ratings, “thumbs up” (recommendation) percentage, hotel class, and chain affiliation. Using a randomized sample of hotels in the Caribbean islands, we employ a three-level mixed-effect linear regression model to investigate the plausible relationship between market accessibility and hotel prices. After controlling for unobserved island-level and hotel-level characteristics, the model indicates that in most periods, low market accessibility (high flight costs) leads to lower hotel prices, and this influence is mitigated by wellestablished positive reputations as represented by the quality-signaling factors. Our findings imply that hotels should work to increase their reputations to help buffer the impacts of inaccessibility. In an effort to increase market accessibility, one course of action is to reduce airport landing taxes and fees. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Hedonic price model Caribbean hotels Market accessibility Quality-signaling factor Mixed-effect linear regression

1. Introduction Market accessibility is defined as the ability of an individual to benefit from a set of opportunities or activities at a destination. This accessibility is determined by the number of opportunities at a place and the cost of realizing those opportunities, which is shaped by several spatial factors (Hanink & Stutts, 2002). Consumers' decisions are shaped by costs of travel, which are related to geographical distance (Nicolau & Mas, 2006). For example, a hotel's location is a fixed attribute. Once established, hotels can hardly move to relocate, therefore consumers must travel to hotels (Yang,

* Corresponding author. E-mail addresses: [email protected] (Y. Yang), nmueller@ufl.edu (N.J. Mueller), [email protected] (R.R. Croes). http://dx.doi.org/10.1016/j.tourman.2016.03.021 0261-5177/© 2016 Elsevier Ltd. All rights reserved.

Luo, & Law, 2014). Mainstream tourism demand studies have defined market accessibility mainly in terms of distance as revealed in transportation costs, and distance is linked to travel costs in tourists' budgets (Peng, Song, & Crouch, 2014). Hotel demand is also linked to travel costs: lower travel costs increase destination access, which leads to increased arrivals and hotel room demand, and consequently, an increase in price. However, to our knowledge, since most hotel pricing studies used the sample of urban hotels without clear information on the origin of hotel guests, the hotel pricing literature has largely overlooked this natural tourism demand process. Search costs are high when products are perishable and intangible (e.g., hotel products), making it difficult for people to gauge product quality (Woodside & King, 2001). Consequently, consumers search for information to reduce uncertainty when purchasing, for example, hotel rooms. Online consumer reviews have

Y. Yang et al. / Tourism Management 56 (2016) 40e51

become important sources of information about the quality and image of a product or service (Gretzel & Yoo, 2008). Consumers consider observable quality signals, such as third-party endorsements (e.g., word of mouth), reputation, guarantees and price, and numerous scholars have investigated how signals affect perceptions of quality and purchasing risk when information is limited (Gretzel & Yoo, 2008; O'Connor, 2010). In the past literature, it is assumed that the effects of quality signaling factors are same across different properties. This one-fit-all assumption is highly problematic due to the great heterogeneity of hotel properties and products they offer. Hence, the relationship between hotel price and quality signals can be contingent upon moderating factors, such as market accessibility. However, these moderators of this relationship has been largely overlooked in the past hotel pricing literature. To fill an important research gap, we investigate the relationship between market accessibility and hotel price using a hedonic pricing framework based on a multi-level dataset from hotels located on Caribbean islands. In particular, we scrutinize the moderating effect of various quality-signaling factors, such as online reviews, chain affiliation, and star rating. This paper is expected make several contributions to the current body of literature. First, this study represents one of the first efforts to investigate the determinants of room prices for Caribbean hotels, as past pricing studies are dominated by urban hotels (Yang et al., 2014). Second, we are particularly interested in the effect of market accessibility on Caribbean hotel prices. Since most Caribbean hotels are categorized as resorts and distant from major markets, market accessibility could play a more substantial role in determining hotel price. Third, although many scholars have incorporated quality signals (espe€g üt & Onur cially online signals) as a determinant of room rate (O Tas¸, 2012; Yacouel & Fleischer, 2012), none have investigated how quality signals may help hotels overcome competitive disadvantages. As information technology becomes more important in shaping customer decision-making processes, understanding the influence of online reputation indicators would help inform marketing strategies and customer service practices. Finally, mainstream hedonic price models do not consider the hierarchy of the dataset nor the unobserved effects stemming from specific factors (Goodman & Thibodeau, 2003; Orford, 2000). In order to address these shortcomings, we employ a multi-level mixed-effect linear regression technique to estimate the proposed hedonic price model. The outputs of this study are particularly useful for practitioners. Understanding the dynamic relationship of these factors has important implications for the role of online consumer opinion platforms in hotel pricing (Zhu & Zhang, 2010). These opinions, to a certain extent, control the strategic considerations of hotel managers. The dynamics affecting tourism price structures in the hotel sector have important economic implications because the sector is a significant driver of local economic benefits due to its vast purchasing power and inputs required to support daily operations (Croes, 2006). The relevance of market accessibility is particularly dominant for island destinations due to natural barriers based on geography, distance and time. Competition on island destinations is condensed to the local level because at a large enough scale, periphery locations do not exist. Almost every inbound visitor planning to stay overnight on an island destination uses air transportation, thereby making air travel costs one of the most important considerations in the decision to visit an island destination. Tourism demand expansion drives economic growth in the short term, and ultimately in the long term as well (Croes, 2003). Market accessibility is thus an overriding and constant concern for Caribbean island destinations and for small island destinations in general, where tourism is an economic driver and the lodging

41

industry is a prevalent catalyst for continued economic growth and security (Croes, 2011). 2. Literature review Hedonic pricing theories were advanced in the mid-1970s by pioneers such as Rosen (1974) and Goodman (1978) in efforts to understand the value of embedded product attributes or characteristics according to revealed market behavior. Based on Lancaster's consumer theory, this modeling strategy assumes that a product's price can be specified as a function of its associated immanent utility-bearing characteristics or attributes (Thrane, 2005). Therefore, a hedonic price strategy enables the total price of a particular good to be disaggregated into separate implicit prices that are attributed to certain inherent attributes or characteristics of the good based on the utility consumers can perceive (Dwyer, Forsyth, & Dwyer, 2010); these implicit prices basically reflect consumer willingness-to-pay (Goodman & Thibodeau, 2003). This method of determining price has been applied to a wide range of products and services that can be decomposed into different attributes with their own implicit prices in equilibrium. The observed price reveals information about potential customers' underlying preferences for those attributes and how businesses can increase the price by including particular characteristics. Empirically, the hedonic price model regresses observed prices on a set of structural, perceptual, and transaction-related factors, and the regression coefficients provide vital information about consumers' willingness-to-pay. Typically, a positive willingness to pay indicates a positive contribution to the level of utility perceived by consumers. Papatheodorou, Lei, and Apostolakis (2012) conducted a thorough review of previous hedonic price model applications in tourism and hospitality management, and they pointed out several advantages of this pricing method, including the ability to price non-market characteristics and the flexibility to accommodate a wide range of pricing factors, In the case of hotels and lodging, the price of a particular hotel's offerings is a function of several attributes of that establishment as well as the local environment. The nature and variety of the specific attributes can be divided into two main categories: internal and external (Chen & Rothschild, 2010). Internal drivers are attributes over which the hotel has control. Attributes that have been considered to impact a hotel's price include, but certainly are not limited to: star rating (Bull, 1994; Espinet, Saez, Coenders, Fluvi, & M., 2003; Israeli, 2002; Thrane, 2007), hotel age (Bull, 1994; Hung, Shang, & Wang, 2010), affiliation with hotel chains/brands (Thrane, 2007; White & Mulligan, 2002), hotel infrastructure (Espinet, Saez, Coenders, & Fluvi, 2003; Thrane, 2005, 2007), the number of rooms available (White & Mulligan, 2002), and the availability of parking (Bull, 1994; Chen & Rothschild, 2010; Espinet et al., 2003; Hamilton, 2007; Hung et al., 2010). External drivers of price are attributes over which the hotel operator has no direct control, such as competition (Balaguer & Pernías, 2013). External attributes are complex and often are considered to have significant influence on pricing as well (Bull, 1994; Chen & Rothschild, 2010; Rigall-I-Torrent et al., 2011; Saleh & Ryan, 1992). Additionally, there is one particular attribute that is unique: location. Although all other internal and external attributes may change, a hotel's location is fixed (Bull, 1994). Location determines proximity to attractions for hotel guests, such as activities or a city center (Chen & Rothschild, 2010; Hung et al., 2010), beaches (Rigall, I-Torrent et al., 2011), or public goods (Rigall-I-Torrent & Fluvia 2011). Moreover, as suggested by spatial interaction theory and the location-allocation model in business geography, a hotel's location relative to another can be captured by market accessibility, such as general accessibility for hotel customers (Fleischer, 2012;

42

Y. Yang et al. / Tourism Management 56 (2016) 40e51

Yang et al., 2014). In the context of hotel demand, spatial interaction theory states that hotel customers trade off the perceived attractiveness of hotel properties against the deterrent effect of travel cost, which can be regarded as market accessibility (Fotheringham & O'Kelly, 1989). In several previous studies, scholars have empirically highlighted the importance of market accessibility in determining tourism and hospitality demand (Hanink & Stutts, 2002; Luo & Yang, 2013; McKercher, 1998). Therefore, it is expected that market accessibility would also play an important role in determining hotel prices, especially on small tourist islands in the Caribbean where hotel supply, at least in the short run, is relatively inflexible. Although a large number of tangible structural factors have been investigated in the previous literature, only a few scholars have examined the effect of quality signaling factors (Abrate, Capriello, & Fraquelli, 2011; Andersson, 2010; Zhang, Ye, & Law, 2011) as a particular type of internal attribute influencing hotel room rates. Quality signaling factors refer to various factors that reduce the information asymmetries in the market by offering buyers information on the quality of products they intend to purchase, and typical quality signaling factors include quality certificates, customer rating, and third-party evaluation. Several theories in information economics explained the importance of these factors during the communication between sellers and buyers. Due to the experience nature of hotel products, the sellers inherently possess more information on product quality than prospective buyers, leading to information asymmetries in the lodging market (Woodside & King, 2001). Akerlof (1970)'s classic “lemons” model highlighted market inefficiency in the presence of severe information asymmetries: if quality cannot be signaled, only lower-quality products will be offered for sale. Therefore, hotel guests increasingly search for various quality-signaling factors online owing to the high search costs associated with the perishable and intangible nature of hotel products (Woodside & King, 2011). Grossman (1981)’s “unfolding” model states that when quality signaling is effective, firms offering high-quality service can get a price premium. Hence, from the supply side, hotels now recognize the importance of establishing and maintaining customer relationships through social media with a low transaction cost; these efforts are reflected in quality-signaling factors and help secure a competitive advantage over peers. Abrate et al. (2011) pointed out the importance of reputationbased quality signals including star rating, brand affiliation, presence in the Michelin Guide, and status of quality assured hotels. Apart from these factors, various online reputation factors have become available with the ongoing penetration of IT technology. For example, some quality signaling factors such as online user rating and “thumbs up” percentage (reflecting the percentage of guests who recommend a hotel) are easily accessible for potential customers via hotel-booking websites. As suggested by customer loyalty theory, electronic word of mouth information about hotels helps potential guests to overcome disadvantages associated with information asymmetry when making booking decisions, and this source of information is believed to be more convincing and reliable than others for gauging the quality of €g üt & Onur Tas¸, 2012). Sparks and Browning (2011) hotels (O showed that a high review rating facilitates hotel booking intentions, and Ye, Law, and Gu (2009) suggested that positive reviews increase the number of bookings for a lodging establishment. Alternatively, if hotel operators are aware of these third-party review websites, it is not hard to imagine room prices being influenced by consumer opinions. According to Grossman (1981)’s “unfolding” model, hoteliers tend to charge price premiums when their establishments have excellent reputations. The effective and

reliable quality signals reduce the search costs of buyers, and at the same time, degrade the uncertainty of purchase, making buyers willing to pay a higher price for the products with positive quality signals. Therefore, it is reasonable to expect a positive association between reputation signals and hotel price. Andersson (2010) unveiled the significant impact of online ratings for facility and food-and-beverage quality in explaining the transaction room prices of Singapore hotels. Zhang, Ye, and Law (2011) evaluated the impact of online reviews on hotel price, and found that the rating scores on room quality and location significantly influence hotel prices. Schamel (2012) included a comprehensive multi-score online rating in the hotel hedonic model from various search engines, and the empirical results confirm that guests are willing to pay a higher premium for a higher rating. Yacouel and Fleischer (2012) proposed a theoretical model showing that a hotelier can charge a premium when a high level of service quality is revealed. They also empirically validated that guests pay a premium when a hotel's service quality has a high rating on travel agent websites. Using a sample of hotel room prices in Paris and €g üt and Onur Tas¸ (2012) found that online customer London, O ratings are positively associated with hotel price and highlighted that the influence of these ratings is different across different types of hotels: room rates for upscale hotels are more sensitive to star ratings. As shown in previous hotel pricing literature, although various quality-signaling factors have been incorporated directly into the hedonic price model, few (if any) researchers have considered the moderating effect of quality-signaling factors on other pricing determinants. The impact of quality signaling effect can be moderated by location-related factors. Gao, Gopal, and Agarwal (2010) found that the return on third-party quality signals is heterogeneous for the Indian software services industry, and this return is dependent on a vendor's location strategy. They showed that the value from a quality signal is particular higher for firms distant from industrial clusters because this signal is effective to reduce the pernicious effects of information asymmetry in a highly uncertain environment related to locational disadvantages. Likewise, McDevitt (2011) indicated that quality signals are more important for firms located in small markets than those in large markets because word of mouth diffuses more rapidly in a small market. From a statistical point of view, most researchers who investigated determinants of hotel prices in the past applied a hedonic price model based on simple linear regression, which manifests a number of shortcomings that compromise the understanding of hotel price dynamics, such as non-stochastic effects and endogeneity. In addition, mainstream hedonic price models do not consider the hierarchy of the dataset nor the unobserved effects stemming from specific factors (Goodman & Thibodeau, 2003; Orford, 2000). In order to address these shortcomings, in this study we employ an innovative multi-level mixed-effect linear regression to estimate the proposed hedonic price model. Scholars have largely overlooked that market accessibility may be a key driver of market prices in the hotel industry (Abrate et al., 2011; Zhang, Zhang, Lu, Cheng, & Zhang, 2011). The specified mixed effects mimic the inter-dependence of error terms in a more straightforward way than other sophisticated models, such as the state space model (Kaidou, Moore, & Charles-Soverall, 2014), the geographically weighted regression model (Zhang et al., 2011), and nez, Sua rez-Vega, & the spatial econometric model (Santana-Jime ndez, 2011). Herna 3. Research design and model specification We considered Caribbean islands readily accessible to the U.S. market. Cuba, the largest island in the Caribbean, was excluded due

Y. Yang et al. / Tourism Management 56 (2016) 40e51

to travel restrictions for U.S. citizens. Hispaniola and Saint Martin were excluded because the existence of more than one country on these islands may have introduced unwanted effects on hotel prices. Several remaining islands were excluded because of their large size or because English is not a commonly spoken language, thus impeding visitation from American tourists. Exceedingly small islands were also eliminated due to concerns over the number of hotels available. From the pool of remaining destinations, we selected ten islands with a wide geographic distribution and considerable heterogeneity in terms of accessibility and levels of tourism development: New Providence, Grand Cayman, Providenciales, St. Thomas, Aruba, St. Vincent, St. Kitts, Grenada, St. Lucia, and St. Barthelemy. The tourism industry for most of these islands depends heavily on the U.S. market. According to the 2013 UNWTO Yearbook of Tourism Statistics, the U.S. market share for inbound tourists exceeds 40% for seven out of the ten islands. The locations of these islands are shown in Fig. 1. Using the website www.TripAdvisor.com, we obtained a list of hotels that included hotel class, user rating, and thumbs up percentage (O'Connor, 2010). For consistency, individual websites for each island or nation were avoided as they utilize different internet travel search services. For each island, we used a random number generating software to randomly select ten hotels among all hotels with user reviews on TripAdvisor.com. This created a set of 100 hotels. To measure the dependent variable, we used the hotel room price for a one-week stay, a common length of stay for our selected islands (Caribbean Tourism Organization, 2010). Hotels often charge higher rates for immediate occupancy, so the one-week stay was selected far enough into the future to avoid this phenomenon (Rigall-I-Torrent et al., 2011). Further, to account for price differences during peak and off-peak seasons (Israeli, 2002; Monty & Skidmore, 2003; White & Mulligan, 2002), we selected three

43

different one-week stays, each in a different month of the forthcoming year relative to the time of data collection. Two of the oneweek stays were selected in the low season outside of the shoulder season months of November and April. (Caribbean low season runs from May through October and corresponds to summer in the Northern Hemisphere and tropical cyclone season in the Caribbean.) The single high season rate was selected in the winter months when American travel to the Caribbean is most pronounced at times that did not coincide with major United States holidays such as Christmas and New Year's Eve. These high and low season patterns are easily observed in tourist arrival data provided by the individual country's statistical report (Caribbean Tourism Organization, 2010). Sunday to Sunday stays were selected to more accurately represent the typical conditions for an average American vacationer who would take a week off from work. The three one-week stays were: June 3e10, 2012 (low season); September 9e16, 2012 (low season); and January 13e20, 2013 (high season). For each of the 100 hotels, the least expensive available price for double occupancy was obtained for each of the three weeks, resulting in 300 initial observations. (The most expensive prices were not collected because there are no limits for luxury.) The total cost of a stay was collected either via automatic online price checking and availability functions for each of the individual hotels, or calculated manually based on published rates on the hotels' websites along with all known taxes and fees. In a handful of cases, we communicated with hotel booking personnel via telephone or email to inquire about missing or expired rates. If the selected week's stay was not available, the prior or following week was used. A price for one hotel was discarded because all of the least expensive accommodations were booked, and those prices were unavailable. All hotels on the island of St. Barthelemy were also

Fig. 1. Location map of the sampled Caribbean islands.

44

Y. Yang et al. / Tourism Management 56 (2016) 40e51

eliminated, because half the hotels from this island were not available for booking in September and only 15% of its tourists arrived from the United States. One hotel in Grenada was excluded as well due to a similar data availability problem. To reflect the hierarchical data structure embedded in our research design, we performed a three-level mixed-effects linear regression to estimate the hedonic price model. There are several noticeable advantages of applying the mixed-effect linear regression model in this study. First, the model takes full advantage of the hierarchical structure of our dataset and accounts for the unobserved hotel- and island-specific factors that influence hotel price but have not been incorporated in the model, such as location relative to major tourist attractions, tourist density, and infrastructure to support tourism. As suggested by several applications of mixed-effect linear models on real estate pricing, overlooking the hierarchical data structure may lead to severe estimate biases (Cervero & Kang, 2011; Goodman & Thibodeau, 2003; Orford, 2002). Second, the model provides an improved estimate to overcome the violation of observation independence assumption embedded in traditional simple linear regression. Compared to other sophisticated models utilized to rectify this problem, the specified random effects mimic the interdependence of error terms in a more straightforward way (Orford, 2000, 2002). We have multiple observations for each hotel over time, and the individual hotels are located on islands. Hence, a three-level hierarchical structure is embedded in this data set. The first-level observation is the hotel price at a particular time, the secondlevel observation is the individual hotel, and the third-level observation is the island where the individual hotel is located. This mixed-effect regression model is specified as follows:

yijt ¼ a þ

p X

bm xm ijt þ mi þ nij þ εijt

m¼1

where t indicates the period of observation (the first level observation), and t ¼ 1, 2, 3; j indicates the hotel (the second-level observation) to which the first-level observation belongs; i indicates the island (the third-level observation) where the hotel is located; a is a constant; and x is a set of p explanatory variables. Apart from the usual error term in a linear regression model, εijt , the model also incorporates two additional random effects: mi denotes the island-level (third-level) random effect that captures unobserved characteristics of the island, whereas nij denotes the random effect of the hotel-level (second-level) random effect that captures unobserved attributes of the hotel. These random effects are not directly estimated, but are summarized based on their estimated variances. Moreover, mi , nij , and εijt are assumed to follow an independent normal distribution with a mean of 0 and an unknown variance. To estimate the proposed mixed-effect linear regression model, we utilize the full maximum likelihood estimation with the expectation-maximization algorithm. This method generates robust estimates that are asymptotically efficient and consistent (Hox, 2010). Based on past literature and data availability, we first specify four quality-signaling factors as independent variables to explain room prices for Caribbean hotels:
rating: The hotel's online user rating (on a scale from 1.0 to 5.0 with increments of 0.5) from a common travel review website, TripAdvisor.com, which captures the electronic word of mouth effect and reputation (O'Connor, 2010).
thumbs: The thumbs up percentage for the hotel from the TripAdvisor website, which reflects the percentage of guests who

recommend the hotel. Unlike the user rating based on a fivepoint scale, users face two options: thumbs up or thumbs down (Pekar & Ou, 2008).
class: The hotel's class based on the TripAdvisor website. This variable ranges from one to five stars, from low to high (Ghose, Ipeirotis, & Li, 2012).
chain: A dummy variable indicating if the particular hotel was part of a known chain. In our model, chain ¼ 1 indicates a chain affiliation, and chain ¼ 0 otherwise. Chain-affiliated hotels are expected to provide standardized products and services to guests, and their quality assurance policies enable guests to reduce potential risks associated with a first-time stay (Abrate et al., 2011). We designated a hotel as being a part of a chain if the parent company held more than 30 individual hotels to ensure brand familiarity among most U.S. customers. As suggested in prior literature, we incorporated other independent variables into the hedonic price model:
lnaccess: Calculated as log(1/cost), where cost is the average cost of a round trip flight (in USD) from New York, Chicago, and Los Angeles to the island where the hotel is located.
period: A nominal variable showing the week associated with the room price (Rigall-I-Torrent et al., 2011); period ¼ 1 for the week June 3e10, 2012; period ¼ 2 for the week September 9e16, 2012; and period ¼ 3 for the week January 13e20, 2013.
beach: A dummy variable indicating access to a private or shared beach (Espinet et al., 2003; Rigall-I-Torrent et al., 2011); beach ¼ 1 indicates access, and beach ¼ 0 otherwise.
business: A dummy variable indicating the presence of a business center on site (Chen & Rothschild, 2010); business ¼ 1 indicates the presence of a business center, and business ¼ 0 otherwise.
fitness: A dummy variable indicating the presence of a fitness center on site (Andersson, 2010; Chen & Rothschild, 2010); fitness ¼ 1 indicates the presence of a fitness center, and fitness ¼ 0 otherwise.
breakfast: A dummy variable indicating the availability of free breakfast on site (Chen & Rothschild, 2010; Fleischer, 2012; Lee & Jang, 2011; White & Mulligan, 2002); breakfast ¼ 1 indicates the availability of free breakfast, and breakfast ¼ 0 otherwise.
internet: A dummy variable indicating the availability of free high-speed internet on site (Chen & Rothschild, 2010; Lee & Jang, 2011); internet ¼ 1 indicates the availability of free highspeed internet, and internet ¼ 0 otherwise.
pool: A dummy variable indicating the presence of a swimming pool on site (Andersson, 2010; Chen & Rothschild, 2010; Rigall-ITorrent et al., 2011; White & Mulligan, 2002); pool ¼ 1 indicates the presence of a swimming pool, and pool ¼ 0 otherwise. Data for these amenity variables were collected from the hotels' websites. Market accessibility in this study is measured by the cost of a flight from major source markets to the island. One could reason that people who can afford a higher travel cost could also afford a higher cost for a week's lodging. Therefore, high flight costs would correlate with higher hotel rates. But, it could also be argued that the increased flight costs would drive the cost of lodging down, owing to the fact that higher transportation costs would decrease demand for lodging and therefore, room costs. We collected data for flight cost (cost) using an online search function, and used the cheapest published rates for available flights coinciding with each of the week-long stays. We selected the three largest cities in the United States (New York, Chicago, and Los Angeles) to offer a range of points of departure.

Y. Yang et al. / Tourism Management 56 (2016) 40e51

45

Table 1 Area, visitor arrival, and accessibility for selected small Caribbean islands. Island

Area (km2)

Air Passenger and (cruise) arrivals

Accessibility (Average air fare in US dollar)

New Providence Grand Cayman Providenciales St. Thomas USVI Aruba St. Vincent St. Kitts Grenada St. Lucia

207 197 98 81 180 340 168 344 620

1,344,189 (4,161,269) 309,091 (1,401,495) 264,877(N/A) 678,692 (2,008,991) 871,316 (599,893) 73,866 (88,925) 106,408 (247,393) 118,295 (309,574) 312,404 (630,304)

837.69 965.77 1138.8 876.46 1003.55 1686.29 1414.9 1803.73 1415.22

Notes: The Air passenger and Cruise arrival data were taken from http://www.onecaribbean.org/statistics/individual-country-statistics. Airfare search services (E.g. Orbitz) was used to quote prices for a budget traveler by selecting the least expensive price quotes, regardless of airline, number of stopovers, or departure times.

Descriptive statistics for the dependent and independent variables are presented in Table 2. We ran several models in order to gain a better understanding of how the independent variables relate to the dependent variable. After preliminary modeling efforts, we decided to use a log-linear model because it outperformed other function forms (Hamilton, 2007; H. Zhang et al., 2011). The dependent variable is lnprice, which refers to the log of total cost (in USD) of a seven-night stay in a hotel. The average user rating for the sampled Caribbean hotels is 4.022, corresponding to a relatively high level of customer satisfaction. The average thumbs up percentage is around 80% and the average hotel class is around 3.5 stars. Only 16.0% of the hotels in the sample are affiliated with a hotel chain, and 53.4% have beach access. In the sample, 50.7% of the hotels have a business center on site, and 50.7% have an on-site fitness center. A free breakfast is available in 25.4% of the hotels, free high-speed internet is offered by 57.5% of the hotels, and a swimming pool can be found at 85.4% of the hotels. Severe multi-collinearity problems will render questionable coefficient estimates in hedonic price models (Andersson, 2010). To detect possible multi-collinearity problems, we created a correlation matrix for the independent variables. By calculating their bivariate correlations, we were able to identify the variables contributing to this problem. As expected, the three qualitysignaling variables obtained from the TripAdvisor website (rating, thumbs, and class) are strongly correlated, with Pearson correlation coefficients ranging from 0.592 to 0.785. This result warns that they cannot be included in a single model at the same time. For other pairwise correlation coefficients, we did not discover strong multicollinearity. Only two are above 0.4, and they are the correlation coefficients for class and fitness (0.494) and fitness and business (0.416). The correlation matrix of these variables is available upon request.

Table 2 Descriptive statistics of variables. Variable

Obs

Mean

Std. Dev.

Min

Max

lnprice rating thumbs class chain lnaccess beach business fitness breakfast internet pool

268 268 223 214 268 268 268 268 268 268 268 268

7.338 4.022 79.395 3.474 0.160 7.081 0.534 0.507 0.507 0.254 0.575 0.854

0.715 0.533 13.347 0.772 0.368 0.278 0.500 0.501 0.501 0.436 0.495 0.353

6.223 2.500 42.000 1.500 0.000 7.612 0.000 0.000 0.000 0.000 0.000 0.000

9.575 5.000 100.000 5.000 1.000 6.666 1.000 1.000 1.000 1.000 1.000 1.000

4. Results and discussion As noted in the previous section, we ran several models in order to gain a better understanding of how the independent variables relate to the dependent variable. Table 3 presents the estimation results of these models that fit the entire data set and data in different periods. Models 1 and 2 are estimated with the traditional OLS linear regression without mixed effects and the mixed-effect linear regression, respectively. Even though the estimated coefficients of rating are similar across these two models, the coefficients of the non-hotel-specific variables, such as lnaccess and period, are very different. Therefore, the results suggest that overlooking the hierarchical structure of the price data leads to unreliable statistically inference in the empirical pricing model. Only mixed-effect linear regression models are further estimated and discussed to answer our research questions. Models 3e4 include all observations and incorporate thumbs and class, respectively. The three online quality-signaling factors in Models 2 to 4 are estimated to be significant and positive, suggesting that guests are willing to pay more in return for a high level of quality and a low level of uncertainty. We denote ex as an exponential function of x. Their estimated coefficients show that one more point in online reviewer rating, 10% more thumbs up, and one more point in hotel class are associated with hotel price premiums of 53.0% ((e0.425e1)  100%), 21.8% ((e0.197e1)  100%), and 75.4% ((e0.562e1)  100%), respectively. Another quality-signaling factor, chain, is positive but statistically insignificant in all three models, corroborating the results from Israeli (2002). The insignificant estimated coefficient of lnaccess provides little support for the plausible positive relationship between hotel room price and accessibility to major markets. For the nominal variable period, period ¼ 1 is set as the reference group. The results in Models 2e4 show that, in general, room rates for Caribbean hotels are higher in January than in June and September. With regard to hotel facilities and service, three factors are estimated to be significant, namely, fitness, breakfast, and pool, which is consistent with the results from White and Mulligan (2002) and Andersson (2010). To explain the marginal effects of these three variables on room rate, we use the estimated coefficient obtained from Model 1. The results indicate that the presence of a fitness center, the availability of free breakfast, and the presence of a swimming pool contribute to room rate premiums of 42.5% ((e0.351e1)  100%), 18.3% ((e0.168e1)  100%), and 55.4% ((e0.441e1)  100%), respectively. To our great surprise, even though beach access should be a great competitive advantage for Caribbean hotels based on the recreational and esthetic values associated with beaches in Caribbean destinations, access to a beach does not significantly contribute to a high room price, as indicated by the insignificant coefficient of beach in most models, which is inconsistent with the finding from Rigall-I-Torrent et al. (2011). For the

46

Y. Yang et al. / Tourism Management 56 (2016) 40e51

Table 3 Estimation results of mixed-effect regression models. Variable

rating

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Model 7

All (OLS)

All

All

All

Jun

Sep

Jan

0.438*** (0.066)

0.425*** (0.114)

0.451*** (0.120)

0.381*** (0.114)

0.489*** (0.132)

0.0550 (0.204) 0.389** (0.179)

0.0146 (0.164) 0.343 (0.219)

0.115 (0.202) 0.442** (0.201)

0.0674 (0.101) 0.0397 (0.156) 0.380*** (0.085) 0.176** (0.076) 0.0493 (0.089) 0.348*** (0.098) 7.635*** (1.384)

0.180* (0.104) 0.00770 (0.112) 0.297*** (0.094) 0.256*** (0.088) 0.00513 (0.072) 0.273*** (0.080) 7.557*** (1.672)

0.266** (0.128) 0.0689 (0.146) 0.296** (0.124) 0.179** (0.083) 0.0720 (0.102) 0.428*** (0.078) 7.959*** (1.518)

5.09e-10

3.77e-13

2.08e-14

0.568*** 90 171.6 194.1

0.454*** 88 128.7 151.0

0.546*** 90 164.4 186.9

thumbs

0.0197*** (0.005)

class chain

0.0143 (0.091) lnaccess 0.379*** (0.140) period¼2 0.100 (0.079) period¼3 0.207** (0.086) beach 0.170** (0.079) business 0.0332 (0.082) fitness 0.328*** (0.090) breakfast 0.203** (0.090) internet 0.00532 (0.072) pool 0.353*** (0.099) constant 7.594*** (1.003) Random-effect coefficient SD(m) SD(n) SD(3) N 268 AIC 444.4 BIC 487.5

0.0817 (0.129) 0.0651 (0.129) 0.0739*** (0.027) 0.204*** (0.039) 0.162 (0.106) 0.0236 (0.144) 0.351*** (0.089) 0.168** (0.076) 0.00633 (0.098) 0.441*** (0.098) 5.348*** (1.267)

0.0782 (0.122) 0.0198 (0.147) 0.0670*** (0.024) 0.233*** (0.039) 0.180* (0.093) 0.0578 (0.177) 0.442*** (0.121) 0.205** (0.081) 0.0575 (0.092) 0.395*** (0.149) 5.218*** (1.308)

0.562*** (0.075) 0.0981 (0.104) 0.0417 (0.136) 0.0701*** (0.026) 0.245*** (0.040) 0.152 (0.108) 0.111 (0.142) 0.183 (0.120) 0.298*** (0.096) 0.00425 (0.118) 0.166 (0.206) 5.338*** (1.108)

5.54e-12*** 0.505*** 0.199*** 268 183.8 216.1

0.0672 0.462*** 0.191*** 223 129.5 156.7

0.0764** 0.385*** 0.191*** 214 102.2 129.2

(Notes: *** indicates significance at 0.01 level, ** indicates significance at 0.05 level, and * indicates significance at 0.10 level. Robust standard errors are presented in parentheses. SD(m) indicates the standard deviation of the island-level random effect, SD(n) indicates the standard deviation of the hotel-level random effect, and SD(3) indicates the standard deviation of the normal error term.).

estimated random-effect coefficients in the lower panel of Table 3, the standard deviation of the island-level random effect is much smaller than its counterpart, the hotel-level random effect, suggesting that hotel-level differences explain much more of the variation in hotel prices than island-level differences. In Models 5e7, we estimate the mixed effect regression model for room rate in three different periods. To keep our sample as large as possible, these three models include rating as the online qualitysignaling measure. Unlike its counterpart in Models 2e4, lnaccess is estimated to be positive and significant in Models 5 and 7, indicating that market accessibility plays an important role in determining hotel room rates in June and January for Caribbean hotels. Some coefficients are estimated to be stable over periods, such as rating and lnaccess, although they are moderately larger in the high season (Model 7). The variable beach is estimated to be positive and statistically significant in Model 7, suggesting that hotels with beach access charge higher prices during the high season. The beach access contributes to room rates in January with a premium of 30.5% ((e0.266e1)  100%). Moreover, Caribbean hotel customers do not value free high-speed internet service, which corroborates the findings from Schamel (2012). For the random-effect coefficient, the estimated standard deviation of island-level random effects is insignificant, showing that there is no significant betweenisland difference in explaining hotel price after controlling for other factors. Finally, to check the validity of model estimates, a further look at the leverage-versus-squared-residual plot from the model

(when additional random effects are removed) suggests the absence of substantial outliers in the sample (Ruppert, 2004). To unveil the moderating factor of the accessibility-price relationship, in Models 8e11 presented in Table 4, we estimate four additional models with the interaction terms between lnaccess and the four quality-signaling independent variables: rating, thumbs, class, and chain, respectively. It turns out that the interactions between lnaccess and rating, thumbs, and class are statistically significant at the 0.01 level, whereas that the interaction with chain is significant at 0.10 level. These results corroborate the moderating effect of various quality-signaling factors. To further explain the marginal effect of lnaccess after introducing the interaction terms, we plot the average marginal effect of lnaccess in Fig. 2. The graphs visualize the marginal effect of lnaccess on hotels with different TripAdvisor user ratings (upper-left), different thumbs up percentages (upper-right), different hotel classes (bottom-left), and chain affiliation statuses (bottom-right). The results suggest that the influence of accessibility decreases along with increases in online user ratings, thumbs up percentage, and hotel class. Therefore, accessibility exerts a smaller effect on room price for hotels with well-established reputations. We can apply the same argument to explain the bottomeright graph, which shows that accessibility is not an important factor affecting room prices at chain hotels. This is further evidence to support the statement that the room prices at hotels with good reputations are less sensitive to the influence of accessibility.

Y. Yang et al. / Tourism Management 56 (2016) 40e51 Table 4 Estimation results of regression models with interaction terms. Variable

rating

Model 8

Model 9

Model 10

Model 11

All

All

All

All

6.009*** (1.308)

0.425*** (0.112) 0.284*** (0.063)

thumbs class chain lnaccess period¼2 period¼3 beach business fitness breakfast internet pool rating  lnaccess

0.100 (0.118) 3.839*** (0.787) 0.0689*** (0.026) 0.213*** (0.040) 0.161 (0.109) 0.0453 (0.156) 0.302*** (0.083) 0.112* (0.061) 0.0345 (0.091) 0.487*** (0.090) 0.922*** (0.195)

thumbs  lnaccess

0.0835 (0.121) 3.688*** (0.762) 0.0559** (0.022) 0.255*** (0.041) 0.212** (0.085) 0.00289 (0.166) 0.358*** (0.102) 0.192** (0.075) 0.0993 (0.084) 0.481*** (0.148)

3.653*** (0.828) 0.109 (0.111) 2.238*** (0.419) 0.0582** (0.024) 0.262*** (0.041) 0.168* (0.088) 0.0559 (0.165) 0.214* (0.113) 0.242*** (0.094) 0.00234 (0.112) 0.129 (0.267)

0.0438*** (0.009)

class  lnaccess

0.594*** (0.118)

chain  lnaccess constant

31.66*** (5.317) Random-effect coefficient SD(m) 8.79e-12*** SD(n) 0.488*** SD(3) 0.196*** N 268 AIC 172.9 BIC 205.2

5.745* (3.277) 0.102 (0.147) 0.0844** (0.034) 0.192*** (0.043) 0.147 (0.109) 0.0368 (0.150) 0.340*** (0.087) 0.156* (0.080) 0.00690 (0.097) 0.438*** (0.096)

30.57*** (5.157)

20.86*** (2.947)

0.852* (0.472) 5.628*** (1.394)

3.94e-11 0.433*** 0.189*** 223 116.5 143.8

0.0405 0.372*** 0.187*** 214 89.48 116.4

2.57e-12 0.505*** 0.198*** 268 182.1 214.5

(Notes: *** indicates significance at 0.01 level, ** indicates significance at 0.05 level, and * indicates significance at 0.10 level. Robust standard errors are presented in parentheses. SD(m) indicates the standard deviation of the island-level random effect, SD(n) indicates the standard deviation of the hotel-level random effect, and SD(3) indicates the standard deviation of the normal error term.).

The models we presented include a wide range of variables that account for the price of a hotel room in the Caribbean. The most intriguing variable is the role of accessibility in determining a hotel's price, especially in the high season. More accessible hotels, measured in this case using the proxy of lower cost of airfare to the island destination, leads to a premium in the hotel price. These findings reflect the distribution of costs a vacationer is willing to spend. Assuming that a traveler has a fixed budget, a lower flight cost to a location would allow for greater expenditure on other items, such as lodging. Generally, more accessible locations will have a wider appeal to vacationers with different attitudes and economic backgrounds. According to Butler's (1980) tourist area cycle of development, a lack of accessibility restricts the number of visitors. As facilities and access improve, the number of visitors will increase. This is supported by Table 1, which shows that generally, more accessible islands have more visitors. Islands, however, can have a limited number of available rooms for guests. As carrying capacity is reached due to land scarcity, crowding or overuse, the

47

number of tourists to the area will begin to stagnate (Butler, 1980). The economic principles of supply and demand explain the relationship between accessibility and hotel price. As the number of visitors increases due to increased accessibility, demand for available rooms increases. This demand, however, cannot be met immediately with an increase in the supply of rooms for several reasons. The tourist area lifecycle states that during the stagnation phase, new development will occur at the periphery. However, small island destinations that have reached the stagnation phase simply cannot keep expanding outwards. They do not have a peripheral area, and off-island alternatives are geographically too distant to be relevant. Hotel and resort construction is also hampered on small islands that are economically less developed and in the tourist area development stage due to a lack of resources and infrastructure. Accessibility to island hotels is largely dictated by accessibility to the island. Once a visitor has decided to visit a certain island, hotel operators must compete for that visitor. As an island becomes more accessible and consequently has a larger supply of visitors, hotel operators can charge higher rates because demand has increased while bed supply has not. This is particularly noticeable in tourist areas with high seasonality where demand fluctuates throughout the year, but bed supply does not. It does not make sense for hotel entrepreneurs to build accommodations if the primary demand for those rooms is only during peak seasons. Our results provide evidence for seasonal effects of accessibility on hotel prices by showing that accessibility has a greater influence on hotel prices during peak season. The number of beds is fixed, yet demand increases during peak season, contributing to the increase in hotel prices. Less accessible islands are subject to the same pressures, but since there is lower overall demand for rooms due to the reduced number of visitors, higher hotel prices are not warranted. This is particularly important for any hotel on a developing island. Efforts to increase accessibility to the island should translate into higher demand for the hotel throughout the year. 5. Conclusions Using a three-level mixed-effect linear regression model, we found that prices charged by Caribbean hotels are influenced by the level of market accessibility, online quality-signaling factors, hotel class, availability of free breakfast, and the presence of a fitness center and a swimming pool. This study contributed to the current literature by introducing the concept of market accessibility into the hedonic pricing model of hotel room rate. Past literature only considered accessibility to attractions (Chen & Rothschild, 2010; Hung et al., 2010; Rigall-I-Torrent et al., 2011) and public goods (Rigall-I-Torrent & Fluvi a, 2011) without looking into the accessibility to origin market. Due to the spatial configuration of source market of Caribbean hotels, we were able to clearly quantify the level of market accessibility measured by travel costs. By simultaneously modeling variables at different levels as well as their interactions, the study revealed that hotel prices in the small island destinations in the Caribbean are associated with the reputation of the hotel, which is consistent with results from Andersson (2010), €g üt and Onur Tas¸ (2012), and Z. Zhang, et al. (2011). Moreover, O we looked into the interaction effect between quality signals and market accessibility, and found that price pressure related to inaccessibility is mitigated when a hotel has a well-established positive reputation in the form of positive user ratings on popular travel websites, or is affiliated with a chain. In other words, these hotel room prices are less sensitive to the influence of accessibility implying that hotel reputation is a stronger driver of hotel prices than travel costs. All previous studies assume the identical effect of quality signals on hotel price, and none of past studies look into the

Y. Yang et al. / Tourism Management 56 (2016) 40e51

-2

-2

Marginal effect on lnprice -1 0 1 2 3

Marginal effect on lnprice -1 0 1 2 3

48

3.0

3.5

rating

4.0

4.5

5.0

40.0

50.0

60.0

70.0 80.0 thumbs

90.0

100.0

-2

-2

Marginal effect on lnprice -1 0 1 2 3

Marginal effect on lnprice -1 0 1 2 3

2.5

1.5

2.0

2.5

3.0 3.5 class

4.0

4.5

5.0

0.0

chain

1.0

Fig. 2. Average marginal effect (AME) and confidence interval (CI) of lnaccess with interaction terms.

heterogeneity of this effect (Abrate et al., 2011; Andersson, 2010; €g üt & Onur Tas¸, 2012; Yacouel & Fleischer, 2012; Z. Zhang et al., O 2011). In fact, to our knowledge, this study represents a pioneering research effort to investigate the quality-signaling factor as a moderator of the relationship between accessibility and hotel price. This study also contributed to the current literature of hotel price modeling from a methodological point of view. The empirical hedonic pricing models are derived from standard ‘pooled’ linear regression models that assume independently and identically distributed (IID) residuals (H. Zhang et al., 2011). Substantively, these pricing models take any two higher-level units and pool them into one singular population. Surely, this assumption seems unreasonable when the model is faced with variables characterized by dependence over time, thereby biasing the standard errors. Additionally, traditional hedonic pricing models have not been effective in capturing unobserved higher-level effects, such as regionspecific and destination-specific effects (like market accessibility, tourist density, and infrastructure quality), that may influence hotel prices. This study alleviated the problem by embedding the hierarchical structure in the research question, as well as by applying a hedonic pricing model based on a multi-level mixed-effect linear regression technique. Our results found that this new technique avoided plausible unreliable statistical inference and outperformed the ‘pooled’ regression in terms of statistical goodness-of-fit. There are several reasons why the mixed-effect model is superior. First, by considering the unobservable and island-specific attributes, the model considered two distinct sources of variance, i.e., between hotels due to differences among destinations and within hotels due

to attributes' changes over time. Second, this technique resolved the issue of endogeneity that could cause and/or result from correlated covariates and residuals in pricing modeling, which incorporate the spatial autocorrelation of pricing data to some nez et al., 2011). extent (Santana-Jime These theoretical propositions have two meaningful managerial implications at both the micro (hotel managers) and macro levels (destination managers). Because customers influence hotel pricing not only through the actual use of the hotel (transaction), but also through social networking, hotel managers are in a better position to influence their hotel prices by enhancing and controlling the reputations of their establishments. Reputation is influenced to a large extent by the ways in which hotel managers stimulate and engage with social media. The other main implication of this study is that hotel managers should constantly and systematically track online reviews and participate in active “conversations” with their customers. In prior studies, researchers assumed a positive relationship between online reviews and hospitality firms' performance and room rates. They focused on the valence and number of lez, Bulchand-Gidumal, & Gonza lez Lo pezreviews (Meli an-Gonza rcel, 2013; Wei, Miao, & Huang, 2013). This study provides the Valca empirical foundation supporting the link between social media (TripAdvisor reviews in this case) and pricing, and demonstrates that social media may affect hotel reputation. Today's customers are able to cast a wider net of influence on other potential customers through the online reviews. The act of posting online reviews influences how hotels set prices, therefore hotel practitioners should actively engage in these online conversations. As suggested

Y. Yang et al. / Tourism Management 56 (2016) 40e51

by the significant moderating effect of online reviews on the accessibility-price relationship, these conversations are particularly important for hotels located on islands with poor accessibility. Consequently, there is a need for hotel managers to systematically monitor online customers' engagement behavior, and to create a system that may facilitate more control over consumers' n-Gonza lez et al. (2013) reviews of their hotel experiences. Melia found that the higher the number of reviews (either positive or negative), the higher the ratings of the reviews. Hotel managers may create direct channels through which customer conversations can be nurtured. One practice could be shifting the review platform away from third-parties (e.g., Trip Advisor) to a hotel's own platform. In this platform, hotels could actively invite customers to share their opinions about their experiences, both positive and negative. Wei et al. (2013) found that positive postings have a greater impact on customers than negative ones, and sharing experiences online increases consumers' perceptions of a hotel's trustworthiness and credibility. Active engagement with customers provides the opportunity for hotels to reveal their earnest efforts to practice service recovery where necessary, and to amplify best practices that resonate with customers. This active engagement not only impacts actual customers, but may also shape the image of the hotel for potential customers who are searching the web for clues regarding the quality of products and services. The credibility of this suggested practice relies, however, on a commitment to deliver a quality hotel experience. The experiential properties embedded in a hotel stay induce a potential moral hazard problem on the hotel's side fueled by the transient nature of service consumption. A hotel manager has an incentive to cheat on quality and thus may not place a high value on reputation. After all, cheating on quality reduces costs immediately, while reputation is associated only with long term outcomes. In addition, building and maintaining reputation is costly and only makes sense when a hotel room can be sold at a premium above its operational costs (Keane, 1996). It appears, therefore, that although a hotel manager must allocate significant resources (money, labor power and time) in order to engage in systematic conversations with customers, the resulting increase in prices should outweigh the costs. In a world of imperfect information and an environment of competition, this may be a tough sell. Achieving and sustaining competitiveness within a high quality context may be a daunting task. Hotel managers may have more incentive to enhance and protect their hotels' reputations when demand for travel to the destination is increased. Because islands operate under a limited demand scheme, more demand means higher hotel prices. More demand can be generated by providing access to more memorable experiences, marketing more aggressively and decreasing travel costs to the destination. Airport taxes and fees create market distortions that potentially affect demand. In many cases, to help recoup the cost of use, airport user fees and landing fees are charged to commercial airlines via contractual agreements (Francis, Fidato, & Humphreys, 2003). The fees imposed upon the airlines are passed along to the lez, Núnez-S n, 2011). passengers (Sainz-Gonza anchez, & Coto-Milla These fees can amount to 25%e40% of the total cost of the airline ticket. For example, a search on Expedia.com revealed that the total cost for a JetBlue round trip ticket from Fort Lauderdale to Nassau on JetBlue for April 13e17, 2014 was $296.50d$165.00 in airfare and $131.50 in taxes and fees. Taxes and fees represent 44% of the total price for this ticket. Since the taxes and fees are fixed amounts, the shorter the flight or the lower the cost of the airline fare is, the greater the proportion of airport fees and government taxes a passenger has to pay. If taxes and fees were eliminated, a round trip ticket from Florida to the Bahamas would cost $199.00, resulting in higher demand for travel to the island. Using JetBlue from New York, the taxes and fees on a flight to Aruba total $114 per person.

49

Without those taxes and fees, the lowest round trip airfare from New York to Aruba would be $341 instead of $455 for February 26eMarch 2, 2014, compared to $358 for a flight from New York to Las Vegas for the same time period. There are two possible ways to eliminate these potential distortions. First, small island destinations could reduce fees unilaterally. During contract negotiations, airport managers and local governments may be able to increase tourism revenue if they reduce airport fees, thus increasing island accessibility. This should in turn increase tourism flows to the island, and result in higher lodging prices. The increased tax flow from more successful hotels and other tourism based companies, may be greater than the revenue lost due to reduced airport fees. Second, under the North American Free Trade Agreement (NAFTA) they could negotiate, for example, with the United States government to eliminate all or a substantial portion of those fees and taxes. This move could be interpreted as removing barriers to their mutual trade. In addition, an agreement forged within the NAFTA framework would also “lock-in” any progress, thereby preventing unilateral revocation if there were preferences, the hallmark of past practices. Endorsing free trade is consistent with the international image of the United States, and supporting tourism in the Caribbean is a good way to compensate for dwindling foreign aid to the region (Croes & Schmidt, 2007). By removing or reducing the taxes and fees on tickets bilaterally with major source markets, transportation costs to the Caribbean would be similar to prices for domestic travel within the United States. Before decreasing airport fees, the tax structure and revenue streams for a specific island should be examined to determine if the reduction in airport fees could be adequately offset by increased tax revenue. For instance, using our data and assuming taxes and fees total 25% of airfare, eliminating these costs would result in an increase in hotel prices, ceteris paribus, by an estimated 9.74% in June (95% CI: 0.97%, 18.50%); 8.58% in the month of September (95% CI: 2.15%, 19.32%), and 11.05% in the month of January (95% CI: 1.22%, 20.88%). The increase in hotel room rates and the commensurate hotel room tax should be compared to the foregone revenues from airport fees and taxes. This process should include stakeholders from the tourism sector who are equipped with necessary information to assess the potential consequences of the tax-demand trade-off. Additionally, these stakeholders may be instrumental in determining how best to use funds for promotional campaigns to increase demand. Although this paper shows how accessibility to small island destinations can influence hotel prices, further research should be conducted to examine if the main theoretical proposition of this study may be generalized to other destinations. The proposition that prices for hotels located on small islands vary directly with reputation and inversely with travel costs is compelling, with meaningful implications for hotel and destination managers. By effectively managing hotel reputation and travel costs, destination managers may be able to alter the competitive landscape in their favor. In the future, researchers should investigate the nature of the relationship between relative island attractiveness and hotel reputation and their interaction, and potential effects on the competitive landscape. In other words, should destination managers pay more attention to hotel reputation or island attractiveness? Is hotel reputation or destination attractiveness consumers' primary consideration? Additionally, scholars should investigate the impact of hotel prices on the sensitivity of different tourist segments to reputation and travel costs. References Abrate, G., Capriello, A., & Fraquelli, G. (2011). When quality signals talk: evidence

50

Y. Yang et al. / Tourism Management 56 (2016) 40e51

from the Turin hotel industry. Tourism Management, 32, 912e921. Akerlof, G. A. (1970). The market for “lemons”: quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 488e500. Andersson, D. E. (2010). Hotel attributes and hedonic prices: an analysis of internetbased transactions in Singapore's market for hotel rooms. The Annals of Regional Science, 44, 229e240. Balaguer, J., & Pernías, J. C. (2013). Relationship between spatial agglomeration and hotel prices. Evidence from business and tourism consumers. Tourism Management, 36, 391e400. Bull, A. O. (1994). Pricing a motel's location. International Journal of Contemporary Hospitality Management, 6, 10e15. Butler, R. W. (1980). The concept of a tourist area cycle of evolution: implications for management of resources. The Canadian geographer, 24, 5e12. Caribbean Tourism Organization. (2010). Individual country statistics. Retrieved November 2014, from http://www.onecaribbean.org/statistics/individualcountry-statistics/. Cervero, R., & Kang, C. D. (2011). Bus rapid transit impacts on land uses and land values in Seoul, Korea. Transport Policy, 18, 102e116. Chen, C.-F., & Rothschild, R. (2010). An application of hedonic pricing analysis to the case of hotel rooms in Taipei. Tourism Economics, 16, 685e694. Croes, R. R. (2003). Growth, development and tourism in a small economy: evidence from Aruba. International Journal of Tourism Research, 5, 315e330. Croes, R. R. (2006). A paradigm shift to a new strategy for small island economies: embracing demand side economics for value enhancement and long term economic stability. Tourism Management, 27, 453e465. Croes, R. R. (2011). The small island paradox, tourism specialization as a potential solution. Saarbrücken. Germany: Lambert Academic Publishing. Croes, R. R, & Schmidt,, P. L. (2007). Promoting tourism as US foreign aid: building on the promise of the Caribbean Basin Initiative. Journal of Interdisciplinary and Multidisciplinary Research, 1, 1e15. Dwyer, L., Forsyth, P., & Dwyer, W. (2010). Tourism economics and policy. Tonawanda, NY: Channel View Publications. Espinet, J. M., Saez, M., Coenders, G., & Fluvi, M. (2003). Effect on prices of the attributes of holiday hotels: a hedonic prices approach. Tourism Economics, 9, 165e177. Fleischer, A. (2012). A room with a viewdA valuation of the Mediterranean Sea view. Tourism Management, 33, 598e602. Fotheringham, A., & O'Kelly, M. E. (1989). Spatial interaction models: Formulations and applications. Dordrecht: Kluwer Academic Pub. Gao, G., Gopal, A., & Agarwal, R. (2010). Contingent effects of quality signaling: evidence from the Indian offshore IT services industry. Management Science, 56, 1012e1029. Ghose, A., Ipeirotis, P. G., & Li, B. (2012). Designing ranking systems for hotels on travel search engines by mining user-generated and crowd sourced content. Marketing Science, 31, 493e520. Goodman, A. C. (1978). Hedonic prices, price indices and housing markets. Journal of Urban Economics, 5, 471e484. Goodman, A. C., & Thibodeau, T. G. (2003). Housing market segmentation and hedonic prediction accuracy. Journal of Housing Economics, 12, 181e201. Gretzel, U., & Yoo, K. (2008). Use and impact of online travel reviews. In P. O'Connor, €pken, & U. Gretzel (Eds.), Information and communication technologies in W. Ho tourism 2008 (pp. 35e46). Vienna: Springer. Grossman, S. J. (1981). The informational role of warranties and private disclosure about product quality. Journal of law and economics, 24, 461e483. Hamilton, J. M. (2007). Coastal landscape and the hedonic price of accommodation. Ecological Economics, 62, 594e602. Hanink, D. M., & Stutts, M. (2002). Spatial demand for national battlefield parks. Annals of Tourism Research, 29, 707e719. Hox, J. J. (2010). Multilevel analysis: Techniques and applications (2nd ed.). New York: Routledge. Hung, W.-T., Shang, J.-K., & Wang, F.-C. (2010). Pricing determinants in the hotel industry: quantile regression analysis. International Journal of Hospitality Management, 29, 378e384. Israeli, A. A. (2002). Star rating and corporate affiliation: their influence on room price and performance of hotels in Israel. International Journal of Hospitality Management, 21, 405e424. Kaidou, D., Moore, W., & Charles-Soverall, W. (2014). Neighborhood features and the rental price of villas and cottages in Barbados. Journal of Hospitality & Tourism Research, 38, 528e545. Lee, S. K., & Jang, S. (2011). Room rates of U.S. airport hotels: examining the dual effects of proximities. Journal of Travel Research, 50, 186e197. Luo, H., & Yang, Y. (2013). Spatial pattern of hotel distribution in China. Tourism and Hospitality Research, 13, 3e15. McDevitt, R. C. (2011). Names and reputations: an empirical analysis. American Economic Journal: Microeconomics, 3, 193e209. McKercher, B. (1998). The effect of market access on destination choice. Journal of Travel Research, 37, 39e47. n-Gonza lez, S., Bulchand-Gidumal, J., & Gonza lez Lo  pez-Valca rcel, B. (2013). Melia Online customer reviews of hotels: as participation increases, better evaluation is obtained. Cornell Hospitality Quarterly, 54, 274e283. Monty, B., & Skidmore, M. (2003). Hedonic pricing and willingness to pay for bed and breakfast amenities in Southeast Wisconsin. Journal of Travel Research, 42, 195e199. Nicolau, J. L., & Mas, F. J. (2006). The influence of distance and prices on the choice of

tourist destinations: the moderating role of motivations. Tourism Management, 27, 982e996. O'Connor, P. (2010). Managing a hotel's image on TripAdvisor. Journal of Hospitality Marketing & Management, 19, 754e772. Orford, S. (2000). Modelling spatial structures in local housing market dynamics: a multilevel perspective. Urban Studies, 37, 1643e1671. Orford, S. (2002). Valuing locational externalities: a GIS and multilevel modelling approach. Environment and Planning B: Planning and Design, 29, 105e127. €g üt, H., & Onur Tas¸, B. K. (2012). The influence of internet customer reviews on the O online sales and prices in hotel industry. The Service Industries Journal, 32, 197e214. Papatheodorou, A., Lei, Z., & Apostolakis, A. (2012). Hedonic price analysis. In L. Dwyer, A. Gill, & N. Seetaram (Eds.), Handbook of research methods in tourism: Quantitative and qualitative approaches (pp. 170e182). Northampton, MA: Edward Elgar Publishing. Pekar, V., & Ou, S. (2008). Discovery of subjective evaluations of product features in hotel reviews. Journal of Vacation Marketing, 14, 145e155. Peng, B., Song, H., & Crouch, G. I. (2014). A meta-analysis of international tourism demand forecasting and implications for practice. Tourism Management, 45, 181e193. , M. (2011). Managing tourism products and destinaRigall-I-Torrent, R., & Fluvia tions embedding public good components: a hedonic approach. Tourism Management, 32, 244e255. , M., Ballester, R., Salo  , A., Ariza, E., & Espinet, J.-M. (2011). Rigall-I-Torrent, R., Fluvia The effects of beach characteristics and location with respect to hotel prices. Tourism Management, 32, 1150e1158. nchez, R., & Coto-Milla n, P. (2011). The impact of Sainz-Gonz alez,, R., Núnez-Sa airport fees on fares for the leisure air travel market: the case of Spain. Journal of Air Transport Management, 17, 158e162. Rosen, S. (1974). Hedonic prices and implicit markets: product differentiation in pure competition. The journal of political economy, 82, 34e55. Ruppert, D. (2004). Statistics and finance: An introduction. New York: Springer. Saleh, F., & Ryan, C. (1992). Client perceptions of hotels: a multi-attribute approach. Tourism Management, 13, 163e168. nez, Y., Sua rez-Vega, R., & Hern Santana-Jime andez, J. M. (2011). Spatial and environmental characteristics of rural tourism lodging units. Anatolia, 22, 89e101. Schamel, G. (2012). Weekend vs. midweek stays: modelling hotel room rates in a small market. International Journal of Hospitality Management, 31, 1113e1118. Sparks, B. A., & Browning, V. (2011). The impact of online reviews on hotel booking intentions and perception of trust. Tourism Management, 32, 1310e1323. Thrane, C. (2005). Hedonic price models and sun-and-beach package tours: the Norwegian case. Journal of Travel Research, 43, 302e308. Thrane, C. (2007). Examining the determinants of room rates for hotels in capital cities: the Oslo experience. Journal of Revenue & Pricing Management, 5, 315e323. Wei, W., Miao, L., & Huang, Z. (2013). Customer engagement behaviors and hotel responses. International Journal of Hospitality Management, 33, 316e330. White, P. J., & Mulligan, G. F. (2002). Hedonic estimates of lodging rates in the four corners region. The Professional Geographer, 54, 533e543. Woodside, A. G., & King, R. I. (2001). An updated model of travel and tourism purchase-consumption systems. Journal of Travel & Tourism Marketing, 10, 3e27. Yacouel, N., & Fleischer, A. (2012). The role of cybermediaries in reputation building and price premiums in the online hotel market. Journal of Travel Research, 51, 219e226. Yang, Y., Luo, H., & Law, R. (2014). Theoretical, empirical, and operational models in hotel location research. International Journal of Hospitality Management, 36, 209e220. Ye, Q., Law, R., & Gu, B. (2009). The impact of online user reviews on hotel room sales. International Journal of Hospitality Management, 28, 180e182. Zhang, Z., Ye, Q., & Law, R. (2011b). Determinants of hotel room price: an exploration of travelers' hierarchy of accommodation needs. International Journal of Contemporary Hospitality Management, 23, 972e981. Zhang, H., Zhang, J., Lu, S., Cheng, S., & Zhang, J. (2011). Modeling hotel room price with geographically weighted regression. International Journal of Hospitality Management, 30, 1036e1043. Zhu, F., & Zhang, X. (2010). Impact of online consumer reviews on sales: the moderating role of product and consumer characteristics. Journal of marketing, 74, 133e148.

Yang Yang (PhD) is assistant professor in the School of Tourism and Hospitality Management at Temple University (School of Tourism & Hospitality Management, 1810 N. 13th Street, Speakman Hall 368, Philadelphia, PA 19122. ). His areas of research interest include tourist flow analysis and financial analysis in the hotel industry.

Y. Yang et al. / Tourism Management 56 (2016) 40e51 Noah J. Mueller is PhD student in the Department of Geography at the University of Florida (3141 Turlington Hall, Gainesville, FL 32611. ). He holds an MS from Purdue University in Interdisciplinary Ecology. His areas of interest include climate change and sea level rise, human-environment interaction in coastal environments and small island developing states (SIDS).

51 Robertico R. Croes (PhD) is professor, Associate Dean, Interim Chair of the Tourism, Events and Attractions department, and Associate Director of the Dick Pope Sr. Institute for Tourism Studies at the Rosen College of Hospitality Management at the University of Central Florida (9907 Universal Blvd., Orlando, Florida 32819. ). His areas of interest include tourist flows, destination management, economic development performance, and quality of life in developing countries and small island destinations.