Using geoinformatics to assess tourist dispersal at the state level

Using geoinformatics to assess tourist dispersal at the state level

Annals of Tourism Research 82 (2020) 102903 Contents lists available at ScienceDirect Annals of Tourism Research journal homepage: www.elsevier.com/...

2MB Sizes 0 Downloads 30 Views

Annals of Tourism Research 82 (2020) 102903

Contents lists available at ScienceDirect

Annals of Tourism Research journal homepage: www.elsevier.com/locate/annals

Using geoinformatics to assess tourist dispersal at the state level ⁎

Anne Hardya, , Amit Birenboimb, Martha Wellsa a b

T

School of Social Sciences, College of Arts Law and Education, University of Tasmania, Private Bag 22, Hobart, Tasmania 7001, Australia Department of Geography and Human Environment, Tel Aviv University, Israel

A R T IC LE I N F O

ABS TRA CT

Associate editor: Sara Dolnicar

Advanced tracking technologies have facilitated the tracking of tourists' movement with high levels of spatial resolution, allowing for the exploration of factors that influence dispersal. However, the degree to which different analytical indices impact the results that they generate, remains under-researched. This study uses a high tempo-spatial resolution data set of location tracking information that was collected in the island state of Tasmania, Australia, using a mobile phone research application. It compares the results that emerge when four different analytical indicators are used to quantify tourist dispersal. The results improve understandings of the role that analytical indicators play in assessing dispersal, along with the underlying factors that influence tourists' dispersal at the state scale.

Keywords: Tourist dispersal Tourist tracking Regional tourism Geoinformatics

Introduction Dispersal of tourists through a destination is currently a pressing topic within new media outlets and tourism destination management organisations. In 2018 and 2019, stories and images of the negative impacts of tourism in Venice, the closing of a popular tourist beach in Boracay and overcrowding on Mount Everest, have made worldwide headlines. The phenomenon of overtourism has been defined (Goodwin, 2017; Milano et al., 2019) and debated (Capocchi, Vallone, Amaduzzi, & Pierotti, 2019) and exposure of the issue has prompted vigorous discussions on the need to disperse tourists into regional areas to decrease congestion, diversify the tourism product and ensure equitable benefits for the tourism industry. This stands in stark contrast to earlier environmentalist approaches that emphasised the negative social and environmental impacts that the dispersal of tourists can have upon remote and regional areas (Fennell, 1996). The goal of moving tourists beyond major tourist gateways and hubs and into regional areas is commonly included in strategic planning documents for tourism agencies, who recognise the potential benefits that tourism may have in terms of economic benefits from expenditure and employment, and social benefits such as pride in one's region. In Australia, for example, the desire and economic necessity for regional dispersal of tourists has resulted in the issue becoming a government priority. The Tourism 2020Implementation Plan 2015–2020 (Australian Trade and Investment Commission (Austrade) and Tourism Australia, 2015; p. 1) lists “Greater regional dispersal by inbound tourists” as its desired outcome. Similarly, regional dispersal is also articulated by the majority of states within Australia, such as Tasmania, where a goal has been set for 66% of all tourist nights in Tasmania to be spent outside the capital city of Hobart (Tasmanian Government, 2019). This Tasmanian goal gives some indication as to how dispersal is defined in Tasmania, and concurs with academic definitions of dispersal by Koo, Wu, and Dwyer (2012). While tourist dispersal within a destination is of major importance, research in the field is limited both in terms of conceptualisation of the term and research methods. Past studies have delved into issues such as flows of international tourist movement



Corresponding author. E-mail addresses: [email protected] (A. Hardy), [email protected] (A. Birenboim), [email protected] (M. Wells).

https://doi.org/10.1016/j.annals.2020.102903 Received 9 August 2019; Received in revised form 25 December 2019; Accepted 6 March 2020 0160-7383/ © 2020 Elsevier Ltd. All rights reserved.

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

(McKercher, Chan, & Lam, 2008; Pearce & Elliott, 1983), and intra-destination movement (Koo et al., 2012; Wu & Carson, 2008). Continuous and precise measurements of distance have rarely been utilised as a method to understand the dispersal of tourists into regions. Moreover, the factors that influence dispersal within a destination have been explored far less frequently. One of the reasons for the dearth in studies of regional dispersal relates to data quality and availability. Lau and McKercher (2006, p. 40) noted that: “The study [of tourist movement within a destination] has been inhibited by the difficulties of gathering useful and detailed itinerary information from tourists”. To measure the exact distance travelled by a regional tourist and the factors that are associated with this dispersal, fine-grained tracking data is paramount. This study utilised high resolution location data and survey data collected through a smartphone application, to investigate the factors that affect tourist dispersal. The study is one of the first and most comprehensive studies of its type to be conducted across an entire state, resulting in tourists' movements being tracked for the duration of their visit, across an entire destination; in this case, it was the island destination of Tasmania, Australia. The outcomes of this study are twofold: first, it introduces new empirically-based conceptualisations for tourist dispersal that up until now were missing due to the scarcity of tourist intra-destination movement information. To do that, the study operationalises and compares four dispersal indicators using geoinformatic tools. Second, it builds upon knowledge of the underlying factors that influence tourists' dispersal and evaluates their relevance for state-level dispersal. Dispersal studies in tourism Early definitions by Cooper (1981) conceptualised dispersal as movement that occurs outwards from a tourism centre to locations with less well-developed tourism facilities. Since this early definition, a variety of others have emerged. Many are built upon the notion of the ‘core’ and ‘periphery’ developed in the 1950s and 1960s by Friedmann (1963) and Myrdal (1957) and then applied to tourism by Christaller (1963). For example, Koo, Wu and Dwyer (2010: 46) defined regional dispersal as “involving at least one night stay in the ‘periphery’” and “…the tendency of visitors to travel beyond the main gateways of the host destination (2010: 1209).” Beyond definitions, a body of work exists in spatial planning that considers spatial distribution, congestion and dispersal. Wall (1997) built upon the work of MacCannell (1976) and Leiper's (1990) tourism attraction system model. Wall (1997) offered an additional, flexible classification system that could be applied to differing scales, ranging from entire countries to single destinations. The author argued that tourist attractions could be classified into three types: 1. point attractions- unidimensional attractions akin to Leiper's notion of clustered nuclei (1990) where large numbers of visitors concentrate in small areas- the attraction is prone to what is now known as overtourism and is essentially the antithesis to dispersal; 2. linear attractions/ resources- two-dimensional attractions that may include lakes and scenic routes, where large numbers congregate along narrow strips of land and the concentration of tourists is not as high as point attractions; and 3. area attractions- multidimensional attractions where large numbers of tourists congregate in more spatially dispersed patterns, such as national parks and wilderness areas. These may include areas of concentration such as visitor centres. More recently, Timothy and Boyd (2015) added the notion of size and reach to understand tourist routes and trails. They argued that trails and routes operate within a nested hierarchy, with the tourist experience at the core and the supply side (including the trails setting, scale and rationale) both playing a role in how tourists disperse along those trails. Practically speaking, different locations appear to prefer different approaches to destination management; some prefer point attractions as a means of directing crowds to specific areas and ensuring that the impacts occur in one location, rather than many locations. This approach stands in contrast to linear and area attractions that involve the dispersal of tourists. Other early research that has explored tourist dispersal has sought to identify itineraries and travel patterns. Mings and McHugh (1992) and Lue, Crompton, and Fesenmaier (1993) conceptualised multi-destination trips and developed a conceptual model of travel patterns, based on destination visited and trip purpose/benefits sought. Lew and McKercher (2006) posited seven different trip patterns of intra-destination movement in their conceptual Travel Dispersal Index (TDI). They argued that factors that influence different travel patterns include transport networks and modes, trip origin and destination, time spent in the destination and personal factors that affect behaviour, plus place knowledge. However, these models of destination choice and dispersal remain largely conceptual. Wu and Carson (2008) posit that empirical understanding of the factors associated with dispersal is a more pertinent consideration, given growth in international travel and focus by governments, who have set targets around tourist dispersal to encourage economic growth. Factors influencing dispersal To date, studies of dispersal and the factors that influence it have been approached from several angles. Broadly speaking these may be classified into: a) factors associated with the traveller (country of origin, life stage, age, travel party composition and size, and reason for visit etc); and b) factors associated with the destination such as transport and the availability of attractions within destinations. These will now be explored. Factors associated with the traveller as influencers of dispersal Early work by Pizam and Sussman (1995) assessed factors such as social interaction, commercial transactions, activity preferences, destination knowledge and the influences that these factors had upon the behaviour of tourists from seven different 2

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

nationalities. The authors argued that of the factors they assessed, differences in their travel behaviour were related to cultural differences. This work concurred with early work by Richardson and Crompton (1988), and while neither of these studies specifically assessed dispersal per se, they can be considered as significant precursors to subsequent studies. Later work by Wu and Carson (2008) concurred. Their exploration of comparative dispersal between domestic and international tourists was conducted using GIS technology and demonstrated that there were differences in dispersal between domestic and international tourists. They furthered the reasoning however, by arguing that dispersal was often the result of the time available to have a holiday in a location. Other traveller factors such as travel career stage, family life stage, physical age of travellers and associated desire to explore have been suggested as factors that may influence travel behaviour and ultimately dispersal (Hardy & Robards, 2015; Prideaux, Wei, & Ruys, 2001). Oppermann (1997a) contended that tourists who stay longer tend to disperse more widely into regions, moving from gateways and major tourists centres first, before dispersing to secondary and tertiary tourism destinations. Motives have also been found to influence dispersal (Le-Klähn, Roosen, Gerike, & Hall, 2015). Debbage (1991) suggested that personality types play a role in identifying the spatial behaviour of tourists. The author found that allocentric-types were more likely to take frequent trips off-island. Conversely, psychocentric tourists tended to restrict their movement to nearby localities. Similarly, Lew and McKercher (2006) argued that allocentric tourists are more likely to travel more widely. More recently, Koo et al. (2012) argued that tourists visiting friends and relatives were strongly associated with higher rates of dispersal. Masiero and Zoltan (2013) argued that tourists with a desire to visit historical places and try new foods were more likely to visit more regions. Fennell (1996) found that special interest tourists consumed a destination quite differently from general interest tourists. They were much more purposeful and directed in their actions and more willing to visit remote, lower-order attractions with a specialist appeal. In addition to motivation and traveller style, aspects such as destination familiarity have also been suggested as factors that influence travel dispersal; early work by Gitelson and Crompton (1984) argued first time visitors dispersed more widely, as did Oppermann (1997b). In a later study Tiefenbacher, Day, and Walton (2000) challenged this by arguing that repeat visitors are more likely to travel shorter distances. Recent work by Lau and McKercher (2004) and Li, Cheng, Kim, and Petrick (2008) has argued that repeat visitors display a tendency to focus their activities; whereas first timers were interested in more general sightseeing. McKercher, Shoval, Ng, and Birenboim (2012) demonstrated that in highly urbanised settings such as Hong Kong, first time visitors tend to focus on iconic attractions, whereas repeat visitors tend to concentrate their visitation to certain activities. There is also a body of work that considers the influence of travel party composition; Tideswell and Faulkner (1999) suggested that while organised tours create repetitive patterns of visitation, the role of the guide is significant- if they choose attractions that are dispersed, regional visitation will occur. Wu and Carson (2008) argued that package tours can play an important role in encouraging dispersal, supported by further work by Koo et al. (2012). Factors associated with the destination as an influencer of dispersal Transport has also been suggested as a major influencer of dispersal. Much focus has been given to the role that the self-drive tourism market may play in encouraging regional dispersal (Hardy, 2003) although in their study of dispersal within Australia, Koo et al. (2010) found that long distance trains and air travel played a larger role in dispersing tourists than self-driven vehicles. Distance, particularly between destinations and attractions, has also been recognised as having a major impact on dispersal. Hwang and Fesenmaier (2003) posited that the resources of destinations, such as accommodation and attractions, also influence the spatial dispersal of tourists to regions. McKercher and Lew (2003) argued that tourists must make trade-offs between their travel time and the time they spend at the destination. As such, as distance increase, demand decreases. This was also explored by McKercher (1998) and Nyaupane, Graefe, and Burns (2003). McKercher and Lau (2008) also showed that territoriality, which they defined as the distance from the hotel, is the main explaining variable of tourist movement patterns. The role of gateways has also been explored; Wu and Carson (2008) illustrated the strengths of Australian gateway cities and the need for differential marketing to encourage international visitors to disperse beyond them. Other aspects specific to the destination, such as weather, has been explored as a destination attribute that affects tourist behaviour. Beckon and Wilson (2013) administered surveys to international tourists at the end of their trips to New Zealand and found that the majority of tourists (63%) made some changes to their trip while travelling, due to weather. Conversely, using a similar methodological approach to this study, McKercher, Shoval, Park, and Kahani (2015) tracked visitors to Hong Kong for one day using GPS technology. They found that the impact of weather was minimal, but noted that tourists to Hong Kong were often on a limited time budget. Traditional approaches to analysing tourist dispersal Descriptive approaches that assess dispersal were popular in the 1980s and are currently experiencing a resurgence in regional tourism organisations whose strategic focus is dispersal. Typically these approaches assess the proportion of number of nights spent in a given destination against the total number of nights spent on the trip (Pearce & Elliott, 1983). The higher the trip index the more important a destination is for a given cohort of tourists; a trip index of 100 means that 100% of the nights were spent in a given destination, and so forth. When applied to regional dispersal, this reliance on the overnight location of tourists dismisses the activity of a tourist that occurs within day trips. Since the development of the trip index, variations have followed. The travel dispersal index (TDI) assessed the overnight location of tourists, but also other trip characteristics, including the length of stay in the country, number of overnight destinations, types of accommodation, the number of transportations used and travel organisation (Oppermann, 1992). More recently, progress has been made through the development of dispersal propensities which allow for a causal analysis of dispersal behaviour (Koo et al., 2012). In 3

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

addition, the Gini coefficient has been developed, which is a measure of statistical dispersal used in economics to depict the distribution of wealth among residents of a country (Keeley, 2015). It is commonly used as a measure of wealth and income inequality, ranging from 0 (perfect equality) to 1 (maximum inequality). Recently, Lau, Koo, and Dwyer (2017) used the decomposition of the Gini index to measure inequalities in the spatial distribution of international tourists in Australia. Recording tourist dispersal using advanced geoinformatics The dispersal indicators that are mentioned above are commonly extracted from census data and retrospective questionnaires which provide a very coarse understanding of tourists' itineraries. Researchers have also utilised activity diaries where participants are asked to give a detailed account of their activities on their vacation including the time and place where the activities took place (see, for example, Fennell, 1996; Thornton, Williams, & Shaw, 1997). This method has several drawbacks: it requires labour intensive and costly coding, is burdensome on participants, is prone to recall bias and it is limited in terms of tempo-spatial resolution and accuracy (Shoval & Isaacson, 2007a). The popularisation of location tracking technologies and geoinformatic tools has offered an alternative to manual recoding of movement. In doing so, they have created new possibilities for analysing tourist time-space movement patterns (Birenboim, AntonClavé, Russo, & Shoval, 2013; Grinberger, Shoval, & McKercher, 2014; Shoval & Isaacson, 2007a). A prominent development in the field are smart phones that contain various built-in tracking technologies and have the ability to store applications (apps) that collect tracking data (Birenboim & Shoval, 2016; Shoval & Ahas, 2016; Shoval, Kwan, Reinau, & Harder, 2014). There are now a variety of options available to researchers who wish to collate fine grained data on tourist's movement. These include the collation of data from mobile phone operators such as log files or call detail records (Nilbe, Ahas, & Silm, 2014; Tiru, Saluveer, Ahas, & Aasa, 2010). While this method has the advantage of providing the researcher with a large dataset, it lacks spatio-temporal precision and frequency of observation points (Shoval & Ahas, 2016). Researchers may also use unique identifiers associated with mobile phones to track movement when mobile phones pass by either a Bluetooth (see Versichele et al., 2014) or Wi-Fi receiver (Hardy et al., 2017). However, this method is limited in its ability to assess movement between two points and is therefore limited in its ability to understand the precise movement of tourists. To overcome this, GPS loggers have often been used to determine tourists' precise movement. App-based research offers the opportunity to produce continuous, fine grained data sets that can determine tourists' movement and dispersal (Birenboim & Shoval, 2016; Hardy et al., 2017). However, previous studies that have attempted to use apps as a research tool have been hampered by participant resistance to participation, perhaps due to concerns over privacy and space (Yun & Park, 2015). At the time of writing, there was a distinct lack of research where apps were used as the research instrument, and a corresponding lack of data that covered entire regions, states or nations (exceptions include research undertaken by Shoval, Schvimer, & Tamir, 2018, and Birenboim, 2018). Dispersal analysis in the age of geoinformatics In this section, we present three alternative approaches for representing and analysing tourist dispersal, namely, maximum distance travelled (a linear approach), ‘total distance travelled’ and ‘activity space’ (an area approach). Each approach requires slightly different data sources and more importantly represents a different understanding of dispersal. It is important to note that there is not one approach that is superior to the others. Moreover, the approaches presented here are not necessarily superior to the more traditional conceptualisation of core-periphery, introduced above. Rather, we argue that each approach is suitable for slightly different purposes. Maximum distance travelled – ‘linear indicator’ Distance is considered to exert a frictional effect on movement (McKercher & Lew, 2003; Yang, Fik, & Zhang, 2017), therefore, it acts as a major travel constraint which affects the magnitude of dispersal. In its most straightforward interpretation, the further a tourist is willing to travel from the core area, the more they are dispersed. Although distance has been recognised as having an impact on the international movements of tourists travelling between countries (McKercher et al., 2008; Nilbe et al., 2014), it has not commonly been used to study regional or state level tourism dispersal (exceptions include Yang et al., 2017). Maximum distance travelled provides a continuous measure of dispersal that does not rely on geographic boundaries and the overnight location of tourists. The use of maximum distance also enriches the interpretation of dispersal; it indicates if tourists have core areas (as with traditional approaches), how far tourists are willing to travel and consequently how many resources the tourist was willing to invest to overcome distance. However, distance as a measure of dispersal is not without flaws. A single measure of distance can be misleading because a single day trip within a week-long holiday could result in a high dispersal index, even if the remaining days of the trip were spent within the boundaries of a city. Alternatively, a tourist who took short day trips to explore regional areas may only be attributed a low dispersal index, despite having spent a very small proportion of their time in the capital city. Total distance travelled – ‘cumulative indicator’ Another variation of a distance or linear measurement is the total distance that a tourist travels throughout the visit. Travel distance is a common measurement in transportation studies where it is used to assess features such as mobility level, accessibility (Geurs & Van Wee, 2004) and impact (Schwanen, Dijst, & Dieleman, 2004). This measurement gives a relatively straightforward account of how much a tourist has travelled in the destination and is useful in understanding the ability and willingness of a tourist to 4

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

overcome space friction. However, travelled distance does not necessarily reflect spatial dispersal as one may travel long distances within the same region without leaving it. Activity space – ‘area indicators’ The concept of activity space is used by geographers to describe the environment one is using and interacting with daily (Golledge & Stimson, 1997). Activity space represents the time-space behaviour of individuals and is a useful tool for representing routine dispersal. It has been adopted in other disciplines, most notably in healthcare and epidemiological research, where it has been used to study the daily mobility of individuals (Jonassaint, Birenboim, Jorgensen, Novelli, & Rosso, 2018) and potential environmental exposure (Perchoux, Chaix, Cummins, & Kestens, 2013). One of the most common operationalizations of activity space is the standard deviation ellipse (SDE) (Rai et al., 2007). SDE represents the spatial magnitude and directionality of dispersal. At one standard deviation, SDE will contain approximately 68% of the activity area (i.e. GPS positions). Here we suggest adopting the concept and operationalization of activity space as an indicator for dispersal in the context of tourism, which is a temporary activity in its nature. Activity space does not indicate how far one is willing to travel to accomplish his or her goals (as is the case with the maximum distance travelled). Rather it characterises the area in which the majority of activities of an individual take place. SDE arguably provides a more solid representation of dispersal. First, in contrast to ratio indicators, it takes into account all travel occurrences and not just overnights. Second, it is not overly sensitive to outliers (i.e. a single distant trip) as is the case with the maximum distance travelled indicator. Third and related to the previous points, activity space indicators such as the SDE take into account the dimension of time. The more time a tourist spends at an attraction the greater the influence of that attraction on determining the magnitude of dispersal. This “temporal feature” is highly relevant to dispersal policy, as we can assume that the more time a tourist spends in a place, the greater their impact (environmental, social-economic etc.) is. Methodologies Data collection The research team decided to utilise a bespoke app, that was designed to collect both GPS data and survey data that was completed after participants consented to take part in the study. The GPS data was recorded every 1 s with 10 m of accuracy. The survey collected data on participants including demographic status, travel party composition and travel behaviour. Participants were recruited at the three major entry points to the island state of Tasmania; Hobart, Launceston and Devonport. In total 1102 groups of tourists were tracked throughout the state, between January 2016 and March 2018. Originally the app was run on a study phone handed out by recruiters (n = 934). Tourists were rewarded for participation with three gigabytes of data and a digitised map of their travel route, upon the completion of their holiday. From late 2017, a stand-alone app was made available to download to non-study phones and tourists were only rewarded with the map of their travels (n = 168). Participants As this study focused on total movement over a complete trip, tourists were only included if a) their tracks contained no time gaps > 24 h; b) they had at least 3 days of data; and c) their first and last points were within 20 km of a gateway port. With these parameters, 38% or 417 tourist tracks were included in the study sample from the total dataset of 1102 tracks. Of the 417 tourists, 396 completed their entry survey. 54% were first time visitors. 69% were Australian. 14% reported that the main reason for their visit was to spend time with friends and family; 86% had other reasons, mostly to see wilderness or wildlife (45%). The age of the participants ranged from 18 to 84 years old, with a mean age of 46 years old. The mean length of stay for all the study sample was 7.4 days. This demographic and trip characteristic data was used to identify factors affecting dispersal so these responses were compared to the 1038 tourists in the total dataset who completed the entry survey. Demographic characteristics such as age, country of origin, income, education and employment status were found to be represented in similar rates in the study sample and the total dataset. First and repeat visitors were also selected in equal rates. There were some distinct differences between the trip characteristics of tourists in the study sample and the total dataset, as tourists selected for the study sample were more likely to: 1) use rented transport (63% compared to 48%); 2) enter the state through Hobart (79% compared to 61%); and 3) have shorter lengths of stay (day counts of three to six days were 49% compared to 38%). Data analysis Following a review of traditional and contemporary approaches to analysing dispersal, the study utilised four indicators that were calculated within a GIS environment using commercial software (ESRI's ArcGIS 10.5.1): 1. Linear (maximum distance) indicator: the study made use of the Euclidean distance between the two most distant GPS points that were recorded for each participant as a measure of the maximum linear distance that was travelled (see illustration in Fig. 1). 5

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Fig. 1. Comparison of dispersal indicators using the case of two tourists' movement within Tasmania.

2. Cumulative distance indicator: The distance that each group of visitors travelled was determined by calculating and adding the distance between temporally adjacent GPS samples. The research app did not record static positions (to improve battery performance) hence, inaccurate positions that are typically recoded when devices are static were avoided this way. 3. SDE (activity space) indicator: An SDE was calculated to represent the activity space of tourists. SDE is commonly used to depict an estimation of the area that is “consumed” by an individual. In addition, the ellipse illustrates the directionality of dispersal (see illustration in Fig. 1). 4. Dispersal ratio indicator: The more traditional dispersal ratio was evaluated by finding the overnight locations of tourists and calculating the proportion of nights they were observed outside Hobart (the main gateway) compared to the total number of nights in Tasmania. Dispersal indicators were compared using a Pearson Correlation Coefficient. While it was expected to find a relatively strong correlation between the various dispersal indicators, some differences were foreseen as well, especially between the ratio indicator and the others. For this project, where the goal was to explore the factors associated with the traveller that were associated with different rates of dispersal, we selected factors that had emerged from previous research and highlighted in the literature review as having an impact upon dispersal. The following demographic and trip characteristics were extracted from the entry survey and GPS data: 1. personal factors: country of origin, travel party composition, transport choice, motivation to visit the destination; familiarity of destination (first versus repeat visitors); time spent in the destination, 2. destination factors: seasonality and entry port. An ordinary least squares (OLS) linear regression was applied on the four dispersal indicators to analyse the impact of personal and destination factors on dispersal patterns. Some of the independent variables that were selected from the entry survey were included despite insignificant values as they were found to be prominent in previous studies. Similar to other studies in the past, the 6

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Table 1 descriptive statistics of the various dispersal indicators. Dispersal indicator statistics

Mean

Median

Min

Max

stdev

Cumulative distance (km) Linear distance (km) SDE area (km2) Dispersal Ratio Days in Tasmania

909.0 178.5 11,454.4 0.44 7.39

806.9 203.1 6169.2 0.50 7

20.1 11.3 0.8 0 3

3878.7 379.5 50,486.2 1 30

604.3 97.6 12,390.6 0.36 3.81

income level was found to be an insignificant predictor of tourist spatial behaviour and was omitted from the models.

Results Fig. 1 illustrates the differences between the dispersal indicators, as calculated for two participants. Both tourists had similar linear distance indicators as both recorded their most distant points near the North Coast and to the South of Hobart. Tourist 1 had a longer length of stay (10 days), travelled a higher cumulative distance (1575 km) and had a higher dispersal ratio (0.889) compared to Tourist 2 (eight days, 939 km, and 0.714), however most of Tourist 1's time was spent close to their gateway, with most nights spent just outside the border used to identify Greater Hobart, so the Tourist 1's SDE is below the sample's average at 6918 km2. Tourist 2's movement was more evenly distributed over their trip, with similar lengths of stay, and multiple overnights, in the South, East Coast and the North, so their SDE is comparatively large. Of the dispersal indicators, the area of the SDE showed the most variance with a minimum value of 0.8 km2 and a maximum value of 50,486 km2, which is 79% of the area of the main island of Tasmania (64,103 km2). Most values, however, were much lower with a median SDE of 6169 km2. The average linear distance was 178.5 km, or similar to the distance between Hobart airport and the popular destination of Cradle Mountain (Table 1). The dispersal ratio indicator showed that an average tourist spent 56% of their nights within Hobart, and 44% in all other locations combined. This dispersal occurred over an average length of stay of 7.39 days. While the linear distance and dispersal ratio had distributions close to normal, the cumulative distance and SDE were rightskewed distributions, with some values many multiples higher than average, but most below average, so in the analysis the square root of both was used. All correlations between dispersal indicators were found to be statistically significant with P values smaller than 0.001 (Table 2). The square root of the SDE and the linear distance were the most highly correlated indicators (r = 0.92). The more traditional Dispersal Ratio indicator demonstrated relatively poorer correlations with the other indicators ranging between r = 0.5 to r = 0.58. Table 3 shows the descriptive statistics of the dispersal indicators for different factors. The results reveal that the number of days spent in Tasmania increased the SDE area rapidly, however, linear distance was less affected by increased day counts. Tourists entering through Devonport had higher levels of dispersal, as did tourists exiting from a different gateway to one's entry. Those Table 2 Correlation matrix of the dispersal indicators using Pearson's R. All relationships were found to be highly significant, with P values < .001.

7

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Table 3 Summary of the travel and their effect on dispersal indicators. Mean results for each factor and (standard deviations) included. Count

SDE area (km2)

Linear distance (km)

Cumulative (km)

Dispersal ratio

Days in Tasmania 3 to 6 7 to 10 11+

195 135 64

6095(8140) 12,954(10,993) 24,788(14,802)

135(84) 206(83) 258(92)

571(328) 1022(413) 1753(694)

0.33(0.36) 0.53(0.33) 0.59(0.29)

Entry gateway Hobart Launceston Devonport

310 36 48

10,044(11,579) 12,080(10,642) 20,315(14,836)

165(96) 195(74) 262(74)

844(547) 956(572) 1367(774)

0.36(0.34) 0.78(0.25) 0.7(0.28)

Exit gateway Same as entry Different to entry

311 83

10,313(12,877) 15,859(9053)

164(100) 239(49)

899(643) 986(427)

0.39(0.37) 0.62(0.23)

Season of trip Spring Shoulder Spring Peak Summer Peak Max Peak Autumn Shoulder Off Season

43 14 166 34 116 21

8053(11,780) 9274(12,476) 13,260(12,385) 12,002(13,048) 11,450(12,404) 5240(9313)

145(107) 160(109) 195(88) 196(104) 177(96) 127(98)

731(501) 615(605) 983(574) 1082(739) 917(617) 726(586)

0.3(0.33) 0.55(0.36) 0.49(0.33) 0.54(0.4) 0.43(0.36) 0.23(0.34)

Transport Own transport Rental transport Buses/coaches Friend/relatives vehicle Other form of transport

52 270 26 33 12

19,100(14,702) 12,212(11,827) 3926(8803) 3189(5978) 19,67(5338)

251(86) 187(87) 90(77) 117(90) 56(77)

1430(816) 932(515) 457(406) 607(391) 263(383)

0.64(0.32) 0.47(0.33) 0.14(0.26) 0.3(0.4) 0.14(0.28)

Companions A couple Group of friends/family Travelling alone

180 172 25

12,598(13220) 11,304(11783) 8743(11,171)

183(101) 185(91) 149(109)

979(655) 905(551) 756(631)

0.44(0.34) 0.47(0.37) 0.37(0.34)

Residence Australia Other overseas country Hong Kong Mainland China

272 90 21 10

10,846(12,663) 12,184(11,723) 12,789(11,521) 20,624(8688)

174(101) 185(86) 192(90) 244(53)

916(656) 904(497) 938(448) 1081(263)

0.44(0.37) 0.45(0.34) 0.37(0.3) 0.53(0.23)

1st time/repeat visitor First visit to Tasmania Returning visitor

212 182

13,420(12,123) 9223(12,304)

196(90) 160(102)

979(574) 847(632)

0.45(0.33) 0.43(0.39)

Purpose Wilderness/wildlife History/heritage Friends or relatives Some other reason Art and culture Food and wine

180 46 55 54 19 39

15,687(13,063) 12,332(12,337) 6946(10,124) 5677(9307) 5052(9438) 8874(9882)

210(86) 185(95) 153(105) 130(96) 108(90) 175(90)

1045(579) 1022(707) 824(673) 671(499) 602(510) 844(492)

0.56(0.31) 0.38(0.32) 0.37(0.38) 0.34(0.4) 0.19(0.31) 0.37(0.33)

traveling alone and during the off-season demonstrated lower levels of dispersal, while transport options were divided with tourists who used their own, or a rental vehicle, dispersing more than other tourists. Table 4 presents the OLS model that was employed with the four independent dispersal indicators. The length of stay in days was found to be the strongest predictor of all dispersal indicators with P values of < 0.001 and longer stays leading to greater dispersal. Repeat visitation was also found to be a strong predictor for dispersal for all indicators except the dispersal ratio, with repeat visitors travelling less than first-time visitors. Entering and leaving from the same gateway also significantly reduced all indicators except cumulative distance. The size and make-up of the group of tourists travelling together was not found to be a major predictor of dispersal, and the season of the trip was only found to significantly impact cumulative distance (reduced distance travelled during spring) and the dispersal ratio indicator (increased dispersal away from the main gateway during peak season). While not significant, all other season coefficients in the model where at the expected direction (positive during peak periods where increased dispersal was expected and negative during low season) with one obvious exception; the SDE coefficient was negative during the peak season. Country of residence was a significant predictor of dispersal for some indicators: The SDE and linear distance were increased when tourists were from Mainland China, however cumulative distance and the dispersal ratio were not. Gateway of entry was also significant for some indicators and not others: entering through Devonport significantly increased the linear distance indicator and entering through Hobart had a strong negative influence on the dispersal ratio. This is to be expected, however as the disperal ratio 8

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Table 4 Multiple linear regression models of the factors affecting dispersal and the four dispersal indicators. Dependent variable:

Entry Gateways

[Launceston] Devonport Hobart

Exit Gateway

[different] Same

Days in Tasmania 1st time/repeat

Residence

[1st time visitor] Repeat Visitor [Hong Kong] Australia Mainland China Other overseas

Companions

[other] Travelling alone A couple Family/friends

Group size Main purpose

[other] Art and culture Food and wine Friends or relatives History/heritage Wilderness/wildlife

Transport

[other] Buses/coaches Friend/relatives Own vehicle Rental transport

Season

[Spring Shoulder] Autumn Shoulder Spring Peak Max Peak Summer Peak Off Season

Constant Observations R2

√SDE

√Cumulative

Linear Distance

Dispersal Ratio

(1)

(2)

(3)

(4)

14.757 (14.096) −6.885 (8.364)

1.407 (1.840) −0.818 (1.092)

43.934⁎⁎ (21.285) −9.498 (12.630)

−0.125 (0.093) −0.374⁎⁎⁎ (0.055)

−26.200⁎⁎⁎ (5.840) 9.700⁎⁎⁎ (0.758)

−0.770 (0.762) 2.014⁎⁎⁎ (0.099)

−50.449⁎⁎⁎ (8.819) 14.152⁎⁎⁎ (1.144)

−0.118⁎⁎⁎ (0.038) 0.026⁎⁎⁎ (0.005)

−17.078⁎⁎⁎ (5.516)

−2.493⁎⁎⁎ (0.720)

−25.181⁎⁎⁎ (8.328)

−0.025 (0.036)

−7.243 (10.623) 43.333⁎⁎ (18.038) −2.660 (10.934)

−1.821 (1.387) 3.616 (2.355) −1.461 (1.427)

−11.093 (16.041) 50.289⁎ (27.237) −3.018 (16.509)

0.038 (0.070) 0.136 (0.119) 0.068 (0.072)

−3.950 (16.123) −9.685 (13.571) −6.634 (12.795) −0.985 (2.569)

−0.469 (2.105) −1.200 (1.772) −0.539 (1.670) 0.086 (0.335)

−14.117 (24.346) −23.053 (20.492) −12.151 (19.320) −2.794 (3.879)

0.036 (0.106) −0.004 (0.089) 0.016 (0.084) 0.007 (0.017)

10.074 (12.213) 18.642⁎ (10.058) 9.538 (9.668) 20.634⁎⁎ (9.312) 37.042⁎⁎⁎ (7.627)

0.996 (1.594) 2.515⁎ (1.313) 0.689 (1.262) 2.622⁎⁎ (1.216) 3.636⁎⁎⁎ (0.996)

−3.357 (18.442) 34.260⁎⁎ (15.188) 15.812 (14.599) 23.410⁎ (14.061) 43.395⁎⁎⁎ (11.517)

−0.072 (0.080) 0.032 (0.066) −0.031 (0.064) −0.011 (0.061) 0.155⁎⁎⁎ (0.050)

−6.248 (15.623) 12.048 (15.737) 25.401 (18.162) 31.832⁎⁎ (13.438)

1.968 (2.040) 7.703⁎⁎⁎ (2.054) 9.056⁎⁎⁎ (2.371) 9.408⁎⁎⁎ (1.754)

5.019 (23.590) 47.160⁎⁎ (23.762) 62.755⁎⁎ (27.425) 72.534⁎⁎⁎ (20.292)

−0.011 (0.103) 0.167 (0.104) 0.085 (0.120) 0.177⁎⁎ (0.088)

7.973 (8.136) −4.016 (14.439) −2.402 (10.621) 5.043 (7.898) −17.045 (11.836) 10.729 (25.152) 386 0.552

0.780 (1.062) −4.079⁎⁎ (1.885) 0.390 (1.387) 0.381 (1.031) −1.893 (1.545) 7.114⁎⁎ (3.284) 386 0.682

6.736 (12.285) −20.017 (21.803) 4.892 (16.038) 4.451 (11.926) −24.612 (17.872) 68.240⁎ (37.979) 386 0.566

0.032 (0.054) 0.105 (0.095) 0.138⁎⁎ (0.070) 0.032 (0.052) −0.084 (0.078) 0.351⁎⁎ (0.166) 386 0.387

(continued on next page) 9

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Table 4 (continued) Dependent variable:

2

Adjusted R Residual Std. Error (df = 359) F Statistic (df = 26; 359)

√SDE

√Cumulative

Linear Distance

Dispersal Ratio

(1)

(2)

(3)

(4)

0.520 43.744 17.017⁎⁎⁎

0.659 5.711 29.605⁎⁎⁎

0.534 66.053 17.987⁎⁎⁎

0.343 0.288 8.714⁎⁎⁎

Note: [reference category] *: P < .05 **: P < .01 ***: P < .001.

was based on the number of nights in Hobart. The self-identified main purpose for visiting Tasmania was also a strong determinant: tourists interested in food and beverages were a significant driver of every dispersal indicator apart from the dispersal ratio, and tourists interested in history significantly increased the SDE and cumulative distance travelled. The most significant purpose, however, was an interest in experiencing Tasmanian wildlife and wilderness. All dispersal indicators increased dramatically for this and were highly significant. Renting transport (either a car/4WD or a campervan/recreational vehicle) was also a significant predictor of increased dispersal and tourists using their own vehicle or using a friend or family member's car significantly increased the cumulative and linear distance travelled. As the length of stay (measured by days in Tasmania) was the most significant variable affecting every dispersal indicator, the linear model was run again, normalised by day count. When normalising by day count, different entry and exit gateways became the strongest predictor of increased dispersal, followed by renting transport and residing in Mainland China. Chinese tourists generally had a short length of stay, close to 5 days, so their country of residence was not seen as one of the most influential factor of dispersal in the original regression models. This illustrated that while tourists from China did not stay as long as tourists from other countries, they dispersed more per day than other countries' residents, and were less likely to spend multiple days in one place. The relationship between the dispersal indicators and length of visit (number of days) was explored more closely due to the

Fig. 2. comparison of dispersal indicators and the relationship between day count. Red points are the average results for each day count. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 10

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Fig. 3. Comparison of visualisations of the purpose of visit, first by using the GPS points in the top row and secondly by using SDEs in the second row.

dominance of the length of stay factor. As expected, it was found that the number of days had less of an impact on linear distance. The graphs in Fig. 2 demonstrate how the relationship between days and linear distance (bottom left graph), and days and dispersal ratio (bottom right graph) became weaker as the day count rose, particularly after the 9th day, while the relationship between SDE and cumulative distance and days was closer to linear. The dispersal ratio also showed a tendency for tourists on short trips to spend all nights in one location - the frequency of 0 or 1 ratio was high for short day counts and decreased as the length of stay increased. As the main purpose of visiting Tasmania was the strongest predictor of dispersal after day count, the movement patterns of tourists by purpose was mapped. Tourists were divided into groups who selected “To see wilderness/wildlife and natural scenery” (n = 179) and those who selected one of the other standard answers (n = 159). 11

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Tasmania was then divided into a grid of two km by two km cells and the following measures were employed: a) the percentage of tourists who had at least one GPS point within a cell; and b) the percentage of SDEs that overlapped with a cell (Fig. 3). It was found that mapping the percentage of SDEs in a location produced a visualization of dispersal that was considerably simpler to identify. Both GPS and SDE maps showed that the two groups of tourists were present in similar areas of the state, were most likely to be near Hobart, and visited more distant regions in different percentages; tourists who wanted to experience wilderness were more likely to be found around the East and West coasts, away from the gateways. However, the complexity of the road system and the lack of timebased data in the map of tourists by GPS points obscured the level of difference between the two tourist groups. Mapping the percentage of SDEs showed that tourists who wanted to experience wilderness spent a greater portion of their time in the state dispersing, and less time in the major gateway. The other indicators: cumulative distance, linear distance, and the dispersal ratio were not legibly mappable like this as all three, despite being derived from spatial data, were non-spatial numerical values. Discussion This study introduced four dispersal indicators that can be used to evaluate tourist dispersal. Three of the indicators (linear, cumulative distance and SDE) were found to be highly correlated with one another (0.84 ≤r ≤0.92) which means that they are likely to generate relatively similar results, as did our OLS model. In contrast, the more traditional disperal ratio indicator was more distinct (with correlations ranging between 0.5 and 0.58 with the other three indicators). Further testing is now needed to assess whether the high correlations of the linear, cumulative distance and SDE indicators would remain in different types and scales of tourism environments. While the indicators are highly correlated, each indicator has unique properties which make it suitable for somewhat different purposes and situations. Table 5 summarises the differences between the indicators and suggests the most suitable usage for each of them. Specifically, the dispersal ratio indicator does not require high spatial resolution information and can utilise ‘overnight stay’ survey data that are relatively simple to collect. While this type of information is not adequate for inferring the spatial magnitude of dispersal, the dispersal ratio indicator is a useful tool for understanding the share of economic impact that takes place within (and outside) a core area, as much of tourist expenditure on accommodation, food, entertainment, transport and shopping takes place near, or within, the accommodations area. Linear measurement of maximum distance travelled can be calculated from more detailed survey data that includes information about the main attractions that are visited, as well as overnight stays. As is the case with the disperal ratio indicator, it is relatively undemanding in terms of data collection. The linear indicator is useful for assessing tourists' willingness to travel and overcome space friction. Willingness to travel relies on data that pertains to both the attractiveness of attractions and its accessibility. It should be noted that the indicator itself does not take into account temporal aspects (e.g. travel time and duration of stay), geographical location and direction. As discussed above (see also Fig. 2) the indicator seems to be less significant in describing dispersal in the case of long vacations (~ > 9 days), in a destination with relatively limited area size, as is the case in Tasmania. To accurately generate SDEs, a more detailed account of tourists' itineraries is required. The SDE indicator probably makes the most comprehensive and solid operationalisation of tourist dispersal since: (1) its calculations take into account both temporal and spatial aspects (including directionality) of tourist spatial movement; (2) the indicator is not highly sensitive to outliers; and (3) it gives a useful graphic representation of the activity area of the tourists. Consequently, SDEs are useful for describing and analysing general patterns of dispersal. However, when used alone it does not address specific behaviours such as the maximum distance that tourists are willing to travel, or the sum of mileage that is travelled. The accurate calculation of cumulative distance is the most demanding indicator in terms of the tempo-spatial resolution of data that it requires. As longer travel distances do not necessarily mean greater dispersal (it is possible to travel long distances within a confined area), it is not possible to automatically infer dispersal based on this indicator. However, as found in this study, the cumulative indicator correlates highly with the activity space SDE indicator. Travel distance measurements are highly relevant for transportation planning such as understanding the impact of tourists travelling on the road network, in terms of both spatial distribution and volume. Factors that affect dispersal at the state level Most of the studies that examined tourist spatial behaviour within a destination have focused on the local scale of an attraction (Birenboim et al., 2013), a city (Shoval & Isaacson, 2007b) and in few cases the region (McKercher & Lau, 2008). However, the same factors that affect the spatial behaviour and dispersal of tourists on a state level appear not to have been explored using integrated geo-spatial and survey technology. This study has demonstrated that the factors that influence tourists' dispersal appear to be similar in both smaller scale environments and at the state-level. Specifically, length of stay is one of the most prominent factors to affect both small scale and state-level dispersal. Our examination of the relationship between different dispersal indicators and day count revealed that dispersal indicators respond differently to long and short lengths of stay. Short holidays did not have a great impact on a tourists' decision to visit destinations far from their gateway. Rather, longer stays in a destination were required before tourists moved their activity base to another region (e.g. another hotel), which in turn resulted in a significant increase of the SDE indicator. Longer stays also resulted in larger amounts of dispersal ratios that were neither 0 or 1 (either entirely within the main gateway or entirely away from it). Significantly, this suggests that longer-staying tourists tend to change from being those which display a hub and spoke movement pattern to those that display a circular or touring pattern. This finding represents a significant theoretical contribution of this study. 12

13

The sum of the Euclidean distance between all consecutive points in a track.

The area of a standard deviational ellipse of a tourist's track with time between points used as a weight.

Cumulative distance

Standard Deviation Ellipse (SDE)

Relatively high resolution time-space information (e.g. from detailed activity diaries GPS).

Extremely high resolution time-space information (e.g. from GPS).

Course location data (e.g. from surveys, activity diaries, GPS).

The Euclidean distance between a tourist's two most distant points.

Linear (Maximum distance) travelled

Not too sensitive to outliers (e.g. single long trips)

Takes into account directionality of dispersal.

Includes useful visualization.

Has no marker for location – all movement could be within, or near, gateway. Takes into account both temporal and spatial aspects of activity.

Indicator's “validity” seems to reduce for long vacations. Includes all meandering.

Sensitive to outliers (e.g. long single day trips).

Does not take time spent in locations as a consideration.

No consideration towards location so long as beyond gateway border. Does not account for location and direction of travel and for the complexity of routes between the points.

Includes time element in proportion of nights.

Does not account for daytime activities.

Overnight stay locations (e.g. from surveys, activity diaries, GPS).

The ratio of overnight stays a tourist spends outside major gateways (In Tasmania: Hobart) to total nights of their trip.

Dispersal Ratio

Main characteristics

Required data

Description

Method

Table 5 A summary of dispersal indicators and their relevance to policy.

General dispersal indicator which takes into account magnitude and directionality of dispersal.

Transportation policy.

Evaluating willingness to travel long distances (affected by attractiveness of attractions and their accessibility).

Evaluating (economic) impact outside and within core areas.

Policy type useful

A. Hardy, et al.

Annals of Tourism Research 82 (2020) 102903

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Another prominent factor that crosses indicators is the familiarity with the destination, as reflected in the number of visits to the destination. In the current study first time visitors were more likely to disperse around Tasmania. These findings are again in line with the spatial behaviour of first-time visitors on a local scale. It seems that first-time visitors who are less familiar with the destination choose more dispersed itineraries in order to be able to attend key attractions and experience more of the destination (Kemperman &s Joh, 2003; McKercher et al., 2012). Group characteristics and socio-demographic variables, including income, were found to be non-significant predictors of dispersal except for a minor effect of the country of origin. Mainland Chinese were more inclined to travel a longer distance and spread more widely as reflected in the linear and SDE indicators. However, this cultural difference should be further investigated; Ryan and Huimin (2007) examined hypothetical planned itineraries of Chinese tourists in the United States and found them to be less dispersed compared to the itineraries of their fellow New Zealand students. There are also a few more unique factors to state-level dispersal that the study emphasises. First, the gateway was found to be one of the most important factors that determines dispersal level. Importantly, it was not so much the entry point that determined dispersal, but rather if tourists were entering and departing from the same gateway. Naturally, when tourists enter a destination from one gateway and exit from another, they are required to travel a substantial distance towards the exit gateway. However, the substantial SDE coefficient suggests that, in this case, dispersal did not simply occur due to a single long drive, but rather it is a result of substantial dispersal across the island. Significantly, the greater dispersal of tourists who enter and exit from different gateways as reflected in the significant coefficients in the SDE, linear and dispersal ratio indicators, did not result in greater distance being travelled (cumulative distance). In other words, while tourists were more dispersed when departing from different gateways, they did not travel longer distances. This makes this strategy both environmentally efficient in terms of transportation mileage, along with being spatially more equitable. Entering and exiting from a different gateway is likely to be a result of itinerary planning made by the tourist; however, it could be promoted by policy tools and incentives to encourage such behaviour. Transportation is yet another important factor in the development and management of state-level tourism. Self-driving options such as travelling with friends and family, using one's own transport and even more so, driving a rental car, were found to increase dispersal. It is important to note that Tasmania does not have public trains or an intra-island airplane network. One final factor for policy makers to consider is the motive for visitation. Similar to Fennell (1996), our study shows that special interest tourists (in our case those interested in wilderness) disperse more widely. While policy makers cannot control tourist motives directly, they may promote certain aspects of their destination to affect the image of the destination, as well as promote or restrict access to more remote attractions thus affecting dispersal patterns. Conclusions The dispersal of tourists from gateways into regional areas is now a major priority for many governments and destination management agencies, as it is seen as a mechanism to distribute the benefits of tourism more equitably and reduce negative impacts of overtourism. The geomatics approach used in this study has demonstrated the accuracy of the data and relative ease with which tourist dispersal patterns may be understood in terms of the factors that influence their dispersal. This has significant implications not only for researchers but also for tourism management organisations who wish to track the dispersal of their visitors and develop transportation strategies. Such strategies may focus on the supply side (e.g. dealing with the location of attractions and transportation infrastructure), the demand side (e.g. tourists itinerary planning), or a combination of the two. This study has made a number of theoretical contributions. It has demonstrated that several of the factors that influence dispersal on a small scale also influence dispersal on a state-wide scale. These include length of stay, familiarity of destination and transport. Significantly it has also demonstrated the influence of state level factors on tourists' dispersal and clustering. Gateways –particularly the entry and exit of tourists through differing gateways- play an important role in influencing dispersal in terms of area and linear distance, although they did not necessarily influence the total distance travelled. A further theoretical contribution that this study makes is the finding that longer-staying tourists tend to change from dispersal patterns of hub and spoke movement to those that display a circular or touring pattern. At a state wide level, in order to avoid congestion, consideration must not only be given to where point attractions, linear attractions or area attractions (Wall, 1997) occur, or to factors such as the tourist experience and administration(Timothy & Boyd, 2015). Rather, aspects such as entry and exit gateways, and tourists' spatial behaviour according to their length of stay, should also be considered when seeking to enhance dispersal. Practically, this study has highlighted the implications of using different dispersal indicators. It has demonstrated that some indicators, such as the dispersal ratio indicator, while coarse in detail, are highly useful for data sets that originate from surveys. The linear indicator is useful for assessing tourists' willingness to travel and overcome space friction, but does not produce temporal or directional data. The SDE indicator creates graphically appealing representations of dispersal and while it requires temporal-spatial data, is useful in describing and analysing general patterns of dispersals. The cumulative distance indicator was found to be very demanding in terms of the tempo-spatial resolution data that it requires. If a destination wishes to enhance dispersal, consideration of the seminal spatial planning work of Timothy and Boyd (2015) on how trails may be used to equitably disperse tourists, will potentially play an important role. But importantly, while this study has highlighted the means by which dispersal may be measured and the factors that are associated with it, there is a counter-argument that this approach is not always required. Rather than dispersing tourism, as noted by Wall (1997), some destinations may choose to take a ‘point attraction approach’ whereby they harden sites and concentrate visitation, in order to prevent the spread of tourism and its associated socio-cultural, environmental and economic impacts. Stewart, Glen, Daly, and O'Sullivan (2001) argued that the 14

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

decision whether to disperse, or not, should be made in the context of broader sustainable management policies for destinations and that dispersal approaches are likely to be more effective at addressing sustainable tourism goals. While this study utilised accurate tools to give a detailed account of tourists' dispersal, it is not without limitations. The study implemented a convenience sampling procedure and also enacted strict inclusion criteria in which only participants with high quality GPS sequences were included in the analysis. This may have led to potential bias that should be further examined in future studies, although the alignment of our results with past studies on factors that influence dispersal suggests that bias may not exist. The study may also be limited as the findings may be specific to Tasmania, which is a small island state with a hard limit on the maximum linear distance that can be reached, and destinations that can be visited on day trips from its gateway cities. Dispersal patterns should, therefore, be examined in additional geographical types of states and at the country scale. Methodologically, dispersal indicators may be further developed into more complex indexes and visualisations that represent several dimensions of tourist spatial behaviour and dispersal. This may eliminate the need to use a few different indicators simultaneously. There is also an opportunity to explore the influence of points of interest upon dispersal. For example, the spatial distribution and the attractiveness level of major tourist sites and cities may be an important factor that affects dispersal. The number of points of interests visited may be used as a dispersal indicator by itself, which may further enhance the approach taken in this study. The preliminary analysis that we conducted found it to be highly correlated with the indexes described in this article. Future studies should address dispersal policy more comprehensively beyond the theoretical impacts of dispersalintroduced here. Attempts should be made to translate the recorded dispersal patterns into economic, environmental and social impacts. This approach may be useful as a supportive tool in the process of dispersal policy design. Statement of contribution 1. What is the contribution to knowledge, theory, policy or practice offered by the paper? This study makes three significant contributions to knowledge. First it assesses how different analytical tools influence the emergence of significant dispersal factors. This is a significant contribution to knowledge regarding the application of geomatics to of tourism research. Second, given that many government organisations are placing great emphasis on measuring and enhancing dispersal, this study contributes to policy because it illustrates the impact that the use of different measurement indicators will have upon the results of their data analysis. Third, the study also makes contributions towards previous research into the underlying factors that influence tourists' dispersal at the state scale. 2. How does the paper offer a social science perspective / approach? This study uses data and analytical tools (from geographical information systems) that have been typically used in the sciences and applies it to a human context- the movement of tourists through a destination. Martha Wells is a research assistant at the University of Tasmania. She is a specialist in geographical information systems and quantitative analysis. Acknowledgments This project was funded by Sense T, the Federal Group, State Growth Tasmania, the University of Tasmania and the Tourism Industry Council of Tasmania. References Australian Trade and Investment Commission (Austrade) and Tourism Australia (2015). Tourism 2020-implementation plan 2015-2020. https://www.austrade.gov. au/Australian/Tourism/Policy-and-Strategy/tourism-2020, Accessed date: 14 February 2018. Beckon, S., & Wilson, J. (2013). The impacts of weather on tourist travel. Tourism Geographies, 15(4), 620–639. Birenboim, A. (2018). The influence of urban environments on our subjective momentary experiences. Environment and Planning B, 54(5), 915–932. Birenboim, A., Anton-Clavé, S., Russo, A. P., & Shoval, N. (2013). Temporal activity patterns of theme park visitors. Tourism Geographies, 15(4), 601–619. Birenboim, A., & Shoval, N. (2016). Mobility research in the age of the smartphone. Annals of the Association of American Geographers, 106(2), 283291. Capocchi, A., Vallone, C., Amaduzzi, A., & Pierotti, M. (2019). Is ‘overtourism’ a new issue in tourism development or just a new term for an already known phenomenon? Current Issues in Tourism. https://doi.org/10.1080/13683500.2019.1638353. Christaller, W. (1963). Some considerations of tourism location in Europe: The peripheral regions – Underdeveloped countries – Recreation areas. Regional Science Association; Papers XII. Lund Congress, 12, 95–105. Cooper, C. P. (1981). Spatial and temporal patterns of tourist behaviour. Regional Studies, 155, 359–371. Deabbage, K. (1991). Spatial behaviour in a Bahamian resort. Annals of Tourism Research, 18(2), 251–268. Fennell, D. A. (1996). A tourist space-time budget in the Shetland islands. Annals of Tourism Research, 234(4), 811–829. Friedmann, J. (1963). Regional economic policy for developing areas. Papers of the Regional Science Association, 11(1), 41–61. Geurs, K. T., & Van Wee, B. (2004). Accessibility evaluation of land-use and transport strategies: Review and research directions. Journal of Transport Geography, 12(2), 127–140. Gitelson, R. G., & Crompton, J. L. (1984). Insights into the repeat vacation phenomenon. Annals of Tourism Research, 11(2), 199–217. Golledge, R. G., & Stimson, R. J. (1997). Spatial behavior: A geographic perspective. Guilford Press. Goodwin, H. 2017. The challenge of overtourism. Responsible Tourism Partnership Working Paper No. 4. October 2017, accessed 10 July at https://haroldgoodwin.info/ pubs/RTP'WP4Overtourism01'2017.pdf. Grinberger, A. Y., Shoval, N., & McKercher, B. (2014). Typologies of tourists’ time–space consumption: A new approach using GPS data and GIS tools. Tourism Geographies, 16(1), 105–123.

15

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Hardy, A. (2003). An investigation into the key factors necessary for the development of iconic touring routes. Journal of Vacation Marketing, 9(4), 314–330. Hardy, A. L., Hyslop, S., Booth, K., Robards, B., Aryal, J., Gretzel, U., & Eccleston, R. (2017). Tracking tourists’ trave; with smartphone-based GPS technology: A methodological discussion. Information Technology & Tourism, 17(3), 255–274. Hardy, A. L., & Robards, B. (2015). The ties that bind: Exploring neo-tribal Theory’s relevance to tourism. Tourism Analysis, 20(4), 443–454. Hwang, Y., & Fesenmaier, D. R. (2003). Multidestination pleasure travel patterns: Empirical evidence from the American travel survey. Journal of Travel Research, 42, 166–171. Jonassaint, C. R., Birenboim, A., Jorgensen, D. R., Novelli, E. M., & Rosso, A. L. (2018). The association of smartphone-based activity space measures with cognitive functioning and pain sickle cell disease. British Journal of Haematology, 181(3), 395. Keeley, B. (2015). Income inequality: The gap between rich and poor. Paris: OECD Publishing. Kemperman, A. D., & Joh, C. H. (2003). Comparing first-time and repeat visitors activity patterns. Tourism Analysis, 8(2), 159–164. Koo, T., Wu, C., & Dwyer, L. (2010). Ground travel mode choices of air arrivals at regional destinations: The significance of tourism attributes and destination contexts. Research in Transportation Economics, 26, 44–53. Koo, T., Wu, C., & Dwyer, L. (2012). Dispersal of visitors within destinations: Descriptive measures and underlying drivers. Tourism Management, 33(5), 1209–1219. Lau, P., Koo, T., & Dwyer, L. (2017). Metrics to measure the geographic characteristucs of tourism markets: An integrated approach based on Gini index decomposition. Tourism Management, 59, 171–181. Lau, A. L. S., & McKercher, B. (2004). Exploration versus acquisition: A comparison of first-time and repeat visitors. Journal of Travel Research, 42(3), 279–285. Lau, G., & McKercher, B. (2006). Understanding tourist movement patterns in a destination: A GIS approach. Tourism and Hospitality Research, 7(1), 39–49. Leiper, N. (1990). Tourist attraction systems. Annals of Tourism Research, 17(3), 367–384. Le-Klähn, D., Roosen, J., Gerike, R., & Hall, C. M. (2015). Factors affecting tourists’ public transport use and areas visited at destinations. Tourism Geographies, 17(5), 738–757. Lew, A., & McKercher, B. (2006). Modeling tourist movements: A local destination analysis. Annals of Tourism Research, 33, 403–423. Li, X., Cheng, C., Kim, H., & Petrick, J. (2008). A systematic comparison of first-time and repeat visitors via a two-phase online survey. Tourism Management, 293, 429–438. Lue, C., Crompton, J. L., & Fesenmaier, D. (1993). Conceptualization of multi-destination pleasure trips. Annals of Tourism Research, 20, 289–301. MacCannell, D. (1976). A new theory of the leisure class. London: Macmillan. Masiero, L., & Zoltan, J. (2013). Tourists’ intra-destination visits and transport mode: A bivariate model. Annals of Tourism Research, 43, 529–546. McKercher, B. (1998). The effect of distance decay on visitor mix at coastal destinations. Pacific Tourism Review, 2(3–4), 215–224. McKercher, B., Chan, A., & Lam, C. (2008). The impact of distance on international tourist movements. Journal of Travel Research, 47(2), 208–224. McKercher, B., & Lau, G. (2008). Movement patterns of tourists within a destination. Tourism Geographies, 10(3), 355–374. McKercher, B., & Lew, A. A. (2003). Distance decay and the impact of effective tourism exclusion zones on international travel flows. Journal of Travel Research, 42(2), 159–165. McKercher, B., Shoval, N., Ng, E., & Birenboim, A. (2012). First and repeat visitor behaviour: GPS tracking and GIS analysis in Hong Kong. Tourism Geographies, 14(1), 147–161. McKercher, B., Shoval, N., Park, E., & Kahani, A. (2015). The [limited] impact of weather on tourist behavior in an urban destination. Journal of Travel Research, 54(4), 442–455. Milano, C., Cheer, J., & Novelli, M. (2019). Introduction: Overtourism: An evolving phenomena. In C. Milano, J. Cheer, & M Novelli (Eds.). Overtourism: Excesses, discontents and measures in tourism. Oxfordshire: CABI. Mings, R., & McHugh, K. (1992). The spatial configuration of travel to Yellowstone National Park. Journal of Travel Research, 30(4), 38–46. Myrdal, G. (1957). Economic development theory and underdeveloped regions. London: Duckworth. Nilbe, K., Ahas, R., & Silm, S. (2014). Evaluating the travel distances of events visitors and regular visitors using Mobile positioning data: The case of Estonia. Journal of Urban Technology, 21(2), 91–107. Nyaupane, G., Graefe, A., & Burns, R. (2003). Does distance matter? Differences in characteristics, behaviors, and attitudes of visitors based on travel distance. Paper presented at the 2003 northeastern recreation research symposium, Newtown Square, PA, U.S.a. Oppermann, M. (1992). Intranational tourist flows in Malaysia. Annals of Tourism Research, 19, 482–500. Oppermann, M. (1997a). Length of stay and travel patterns. In R. Bushell (Ed.). CAUTHE 1997: Tourism research: Building a better industry. Proceedings from the Australian Tourism and Hospitality Research Conference (pp. 471–480). Canberra: Bureau of Tourism Research. Oppermann, M. (1997b). First-time and repeat visitors to New Zealand. Tourism Management, 183, 177–181. Pearce, D. G., & Elliott, J. M. C. (1983). The trip index. Journal of Travel Research, 22(1), 6–9. Perchoux, C., Chaix, B., Cummins, S., & Kestens, Y. (2013). Conceptualization and measurement of environmental exposure in epidemiology: Accounting for activity space related to daily mobility. Health & Place, 21, 86–93. Pizam, A., & Sussman, S. (1995). Does nationality affect tourist behavior? Annals of Tourism Research, 22(4), 901–917. Prideaux, B., Wei, S., & Ruys, H. (2001). The senior drive tour market in Australia. Journal of Vacation Marketing, 7(3), 209–219. Rai, R. K., Balmer, M., Rieser, M., Vaze, V. S., Schönfelder, S., & Axhausen, K. W. (2007). Capturing human activity spaces: New geometries. Transportation Research Record, 2021(1), 70–80. Richardson, S. L., & Crompton, J. (1988). Vacation patterns of French and English Canadians. Annals of Tourism Research, 15, 430–448. Ryan, C., & Huimin, G. (2007). Spatial planning, mobilities and culture—Chinese and New Zealand student preferences for Californian travel. International Journal of Tourism Research, 9(3), 189–203. Schwanen, T., Dijst, M., & Dieleman, F. M. (2004). Policies for urban form and their impact on travel: The Netherlands experience. Urban Studies, 41(3), 579–603. Shoval, N., & Ahas, R. (2016). The use of tracking technologies in tourism research: The first decade. Tourism Geographies, 18(5), 587–606. Shoval, N., & Isaacson, M. (2007a). Tracking tourists in the digital age. Annals of Tourism Research, 34(1), 141–159. Shoval, N., & Isaacson, M. (2007b). Sequence alignment as a method for human activity analysis in space and time. Annals of the Association of American Geographers, 97(2), 282–297. Shoval, N., Kwan, M.-P., Reinau, K. H., & Harder, H. (2014). The shoemaker’s son always goes barefoot: Implementations of GPS and other tracking technologies for geographic research. Geoforum, 51(1), 1–5. Shoval, N., Schvimer, Y., & Tamir, M. (2018). Real-time measurement of tourists’ objective and subjective emotions in time and space. Journal of Travel Research, 57(1), 3–16. Stewart, E., Glen, M., Daly, K., & O’Sullivan, D. (2001). To centralise or disperse – A question for interpretation: A case study of interpretive planning in the Breck. Journal of Sustainable Tourism, 9(4), 342–355. Tasmanian Government, 2019. Measuring Progress. https://www.t21.net.au/home/homepage-links/measuring-progress (accessed May 9 2019). Thornton, P. R., Williams, A. M., & Shaw, G. (1997). Revisiting time-space diaries: An exploratory case study of tourist behavior in Cornwall, England. Environment and Planning A, 29, 1847–1867. Tideswell, C., & Faulkner, B. (1999). Multidestination travel patterns of international visitors to Queensland. Journal of Travel Research, 37(4), 364–374. Tiefenbacher, P., Day, F. A., & Walton, J. A. (2000). Attributes of repeat visitors to small tourist-oriented communities. The Social Science Journals, 37(2), 299–308. Timothy, D. J., & Boyd, S. W. (2015). Tourism and trails: Cultural, ecological and management issues. Toronto: Channel View Publications. Tiru, M., Saluveer, E., Ahas, R., & Aasa, A. (2010). The Positium barometer: A web-based tool for monitoring the mobility of tourists. Journal of Urban Technology, 17(1), 71–89. Versichele, M., de Groote, L., Bouuaert, M., Neutens, T., Moerman, I., & Van d Weghe, I. (2014). Pattern mining in tourist attraction visits through association rule learning on Bluetooth tracking data: A case study of Ghent, Belgium. Tourism Management, 22, 67–81. Wall, G. (1997). Tourism attractions: Pints, lines and areas. Annals of Tourism Research, 24(1), 240–243.

16

Annals of Tourism Research 82 (2020) 102903

A. Hardy, et al.

Wu, C., & Carson, D. (2008). Spatial and temporal tourist dispersal analysis in multiple destination travel. Journal of Travel Research, 46(3), 311–317. Yang, Y., Fik, T. J., & Zhang, H. L. (2017). Designing a tourism spillover index based on multidestination travel: A two-stage distance-based modeling approach. Journal of Travel Research, 56(3), 317–333. Yun, H., & Park, M. (2015). Time–space movement of festival visitors in rural areas using a smart phone application. Asia Pacific Journal of Tourism Research, 20(11), 1246–1265. Anne Hardy is Associate Professor at University of Tasmania, Australia. She conducts research in three areas: the neo-tribal behaviour of tourists; sustainable tourism; and tracking tourists' movement. Amit Birenboim is a senior lecturer at the Department of Geography and the Human Environment, Tel Aviv University. His research interests include the implementation of sensors and advanced location tracking technologies to the study of human spatial behavior.

17