2. Theoretical model
Theoretical models establish the theoretical foundation for the spatial location choice of hotels. Theories from different disciplines have been used to explain different perspectives on hotel location. These theories include geographical (Egan and Nield, 2000; Shoval, 2006), economic (Kalnins and Chung, 2004) and marketing theories (Baum and Haveman, 1997; Urtasun and Gutiérrez, 2006). We categorize previously documented theoretical models into four types based on their disciplinary backgrounds, and they are the touristhistoric city model, the mono-centric model, the agglomeration model, and the multi-dimensional model.
2.1. Tourist-historic city model (THC model)
THC models date back to Ashworth and Tunbridge’s (1990) comprehensive typology of hotel locations within medium-sized Western European provincial towns. In their work, six types of location zones were identified, including traditional city gates (A), railway station/approach roads (B), main access roads (C), “nice” locations (D), transition zones and urban periphery on motorway (E), and airport transport interchanges (F). These different zones are associated with different types of hotels. For example, large modern hotels can be found in type E and type F locations, whereas small and medium hotels dominate type D locations. They attributed these clusters to the influence of access, land values, environmental convenience, historical continuity, and land-use policy.
In tourism and hospitality studies, there is a long tradition of applying the THC model to investigate hotel location and spatial distribution in tourist-historic cities. Most tourist cities have been found to exhibit a hotel distribution pattern postulated by the THC model. Burtenshaw et al. (1991) applied the THC model to explain the typology of hotel distribution in several European cities. To interpret hotel evolution from a spatial perspective, Timothy and Wall (1995) studied the accommodation in Yogyakarta, Indonesia and discovered that the THC model can reasonably explain the location of hotels and predict the locational classification of accommodations. Furthermore, Oppermann et al. (1996) used this model to discuss the hotel distribution in Kuala Lumpur, Malaysia. In their study, seven types of location zones were recognized, and the most distinguished was the “new Central Business District location.” This included large modern hotels and deluxe shopping centers, which are common in Southeast Asian countries. Rogerson (2012a) also highlighted the importance of CBD in attracting hotels in three cities of South Africa, and identified some “nice” locations for hotels as described in the THC model.
In another study by Bégin (2000), it was found that hotel locations in Xiamen, China, in general, coincided with those described in the THCmodel. A large number of cheap hotels were clustered in the historical center, and new hotels were constructed in the transition zone between the old downtown and the emerging CBD. Shoval and Cohen-Hattab (2001) investigated the location of tourism accommodations in Jerusalem, Israel over the past 150 years. Focusing on four periods of development, the study confirmed the predictions of the THC model. It also highlighted other important factors shaping hotel distribution, such as political upheavals and social and cultural differences between the population groups. Aliagaoglu and Ugur (2008) found that the results from Dökmeci and Balta (1999) on hotel location pattern in Istanbul, Turkey confirmed the THC model’s prediction, and both type A and type E locations in the city were identified.
The value of the THC model lies in its simplicity and briefness to consider major location hotspots for hotels and the general spatial arrangement within a tourist city. Although it is very popular in the tourism literature, the THC model is subject to many limitations. First, as indicated by Ashworth and Tunbridge (2000), the model is taxonomic rather than explanatory. As such, even though the potential location for hotels within the city can be identified, we do not understand the exact reason why it is selected. Apart from that, while this model has been found to be applicable to touristhistoric cities, it may not be appropriate for non-tourist-historic cities (Aliagaoglu and Ugur, 2008; De Bres, 1994). If it is applicable, however, then, what improvements or modifications should be made to cater to this new situation?
2.2. Mono-centric model
The mono-centric model describes the distribution of land use patterns as several mono-centric rings according to the distance from the city center and emphasizes the paramount importance of accessibility in shaping this pattern (Alonso, 1964; Von Thünen, 1826). In the model, it is assumed that an urban area is monocentric with a single central point for sprawl, and the bid-rent curve is introduced to depict how much land users are willing to pay for locations with different proximities to the center. Based on the principle of bid-rent curves, and drawn from Von Thünen’s (1826) land-use model, Yokeno (1968) proposed a mono-centric model to highlight the possible location of urban hotels. With an assumption that tourists are willing to pay more in return for easy access to the city center, the new model suggested that the hotel district is in the center of the city, located between the city’s innermost CBD and commercial zones (Fig. 2a).
Egan and Nield (2000) derived another mono-centric model from the partial-equilibrium bid-rent approach, and explained the spatial hierarchy of hotels in terms of the distance to the city center. Land bid-rent curves highlight the revenue associated with locations, and it is assumed that hotels’ revenue falls when they move to locations away from the center. In the model, the location preference of hotels of different levels could be predicted by the shape of the bid-rent curve associated with them. Luxury hotels (four-/fivestar) are expected to have a very steep and high bid-rent curve and prefer a central location (Fig. 2b). This is because their higher room rates targeting affluent guests are likely to cover the higher land values associated with a central location. Conversely, due to insufficient revenue to pay for a central location, budget hotels choose to either locate at the edge of the city, or select converted buildings at the edge of the city center. To further validate the generality of Egan and Nield’s (2000) model, Egan et al. (2006) tested hotels in three Chinese metropolises: Beijing, Shanghai, and Shenzhen. Their results suggested that the hotel location in these cities generally fit the model well, despite some minor flaws. Many other cities have been found to contain a spatial hierarchy and concentric arrangement of hotel distribution that is analogous to Egan and Nield’s (2000) model, such as Cape Town, Durban, and Port Elizabeth in South Africa (Rogerson, 2012a) and the Kumasi Metropolis in Ghana (Adam, 2013).
In addition, Shoval (2006) demonstrated that Yokeno’s (1968) model was capable of predicting hotel location in Jerusalem, Israel. He proposed an extended model by recognizing two geographies of demand for hotels: the hotel area for individual tourism and for organized tourism (Fig. 2c). Different markets corresponded to different bid-rent curves. In a more comprehensive empirical study conducted by Yang et al. (2012), the mono-centric model was used to explain this spatial hierarchy of hotel distribution in Beijing, China. Based on the bid-rent analysis, the mono-centric model can also be generalized to study the city with dual centers, and an overlapped spatial hierarchy of hotel locations to each center has been identified (Egan et al., 2006; Lee and Jang, 2011).
In sum, mono-centric models provide a powerful analytical tool, bid-rent analysis, to look into hotel location and other activities within the scope of the whole city. In general, these models highlight a centripetal force on upscale hotel locations while a centrifugal force on downscale ones. Several empirical studies have supported the usefulness of this model in predicting the spatial arrangement of hotels within a city. However, because of the complexity of the hotel location problem, the mono-centric model investigates it under several oversimplified conditions (Shoval, 2006, p. 63), and some of them have been deemed too unrealistic for general hotel location cases. For example, it is inappropriate to assume that the city as a mono-centric one in most situations (Lee and Jang, 2011) and posit the central location as the major or even the sole preference of hotel guests. Moreover, the mono-centric model does not adequately capture all aspects of hotel location patterns, and most importantly, it fails to explain micro-scale hotel agglomeration (Egan et al., 2006), which has been accepted as conventional wisdom.