Estimation of city tourism flows: challenges, new data and COVID

Quantifying attractiveness and city tourist demand

Cities are attraction points for people to visit for many reasons. Recreational, culinary, cultural facilities, historical buildings and monuments are all undoubtedly reasons why people often make expensive travel to these destinations. To enhance the tourism experience, many cities have been aiming to improve the transportation experience. Cleanliness, convenience and maintaining or improving the city’s transportation facilities are all aspects of this. A difference to planning for residents is that aspects such as travel times are to some degree of less importance than wayfinding, information and aesthetic aspects.

Moreover, cities are competing for tourists as their expenditures are for many cities a major source of income. A recent keyword, often highlighted in this discussion and used to distinguish a destination from competing ones, is “authenticity” or the promise of delivering “an authentic experience”. Even if another place would provide the same or an even better experience, being at the “real” place has an intrinsic value itself for tourists (e.g. Park, Choi, & Lee, 2019). Though both recent and more historical events can be the source of authenticity, in general authenticity appears to be linked to time. Therefore, the longer the history the more authentic a place tends to be. Clearly this is one reason why, for example, European cities compared to many American and Asian cities, have an advantage as their architectural history stretches much further back. The values of buildings such as cathedrals for trip attraction during their century lasting life span are nearly impossible to estimate. As authenticity is difficult to create, events are a much faster way to stimulate attraction. Large one-time events such as staging the Olympics, or recurrent events such as fairs or smaller events create visitor streams in the short run. Events are further creating the possibility of some long-term effects as the city gains awareness (as well as, over time, some authenticity).

If one wants to forecast the demand for travel to a city, the number of visitors to events appear to be relatively easy to estimate in some cases, e.g. by the number of tickets sold. In addition, data from restaurants, hotel bookings, etc. can catch trends in tourist visits. Furthermore, the overall number of tourists attracted can be estimated from previous years but will undergo variations. These variations tend to be larger when compared to the travel trends of residents whose patterns are constrained by work, school and other fixed activities. Tourist destination choice lacks these constraints. Furthermore, variations are increasingly triggered by positive and negative feedback loops by various social media and public “ratings” of visiting experiences of past experiences. Some places that are “in” experience a large increase, whereas places that have bad (social) media coverage tend to experience large reductions in flows or need to compensate these trends with price reductions. Including such issues in short term tourist demand forecasting remains an important issue and various new data sources can help with this. In the next section, some data sources are discussed for tourist flow estimation inside a city that can also provide information on total flows. As an example for the relationship between social ratings and demand, Möhring, Keller, Schmidt, and Dacko (2020) have shown that google ratings and the number of uploaded pictures are related to an estimate in tourist numbers.

Even more difficult is to build explanatory models for long-term forecasting and to evaluate the importance of specific facilities. The author lives in Kyoto, Japan, which has 14 UNESCO world heritage sites. Cost–benefit analysis in terms of additional tourist numbers attracted to the city overall given increased maintenance spending on one of these sites is very speculative.

Tourist flow patterns inside the city and data sources

For transport planning the overall number of tourists visiting a city and its temporal distribution is the foundation for then modelling tourists’ flows inside the city. Compared to the estimation of flows for residents the literature appears to be fairly sparse with respect to trip chaining sequences and mode choices of city visitors. In many cities, tourist flows might not be of sufficient importance for the overall traffic conditions but that is not true in tourist–heavy cities. Referring again to Kyoto as an example, both road and public transport throughout the whole city undergo significant seasonal patterns in line with annual seasons that drive tourist demand. In particular the cherry blossom season in spring and the autumn foliage season lead to heavy congestion not just near sightseeing spots but throughout the city.

The foundations for transport planning data in most cities are still household surveys which do not cover tourists. In Kyoto, an add-on survey was conducted to capture tourist behaviour 15 years ago, but this survey was carried out in one particular season and not renewed since then. New data sources are, however, increasingly available and allow us to estimate flows. Mobile phone service providers can distinguish data from those who have locally registered their phone and those from other parts of the country or abroad. Such data can be a powerful source to obtain insights into tourist behavioural patterns depending on the levels of their spatial and temporal resolution and any privacy constraints (Phithakkitnukoon et al., 2015). Alternatively, sensors installed throughout the city that capture Mac addresses of those carrying Wi-Fi or Bluetooth enabled devices provide valuable data (Nunes, Ribeiro, Prandi, & Nisi, 2017). Also here, it appears possible to at least in some cases distinguish the country where mobile phones are registered and hence to identify those who are likely foreign tourists. In our own currently ongoing research, we use such a set of sensors installed at various points throughout Kyoto to obtain estimates of person flows, as well as tour patterns from being able to re-identify some Mac addresses at different sensors.

Another source of locational data are other mobile applications targeted at tourists that might leave locational footprints with user agreement. A local transport planning application in Kyoto, for example, asks users for their consent to record locational records. Booking, tour guidance and other informational apps are further data sources as the frequency of enquiries for specific locations will, to some degree, affect travel patterns. To capture the above-mentioned fast changing trends in tourist destinations, it is also important to note that there are increasingly rich possibilities to analyse social media data. Importantly this data source is rich with tourist information as one is more likely to report on such activities instead of routine activities. Twitter and other media that allow web crawling of user posts can lead to new understanding of tourist travel choice dynamics.

There are also differences between increased predictive power and being able to understand motivations that need to be noted. Papers such as Gavalas, Konstantopoulos, Mastakas, and Pantziou (2014) consider the choice of tour patterns in a city as an optimisation problem from a traveller perspective. However, for both route guidance as well as flow prediction defining the objective function is highly speculative as tourists are one of the most diverse travel groups. Some rush between sites to cover the largest possible number of attractions within their time frame, whereas others are satisfied by visiting one major sight and others are aiming to simply “catch the atmosphere”. These are examples of differences in attitudes to the number of points to be visited during a tourist day even before considering preferences for different type of attractions. Also here behavioural models for residents are not easily transferable as preferences do not only vary depending on personal socio-demographics and attitudes but also on how often persons have been visiting a place already, as well as the specific information tourists might have received during their trip.

COVID and city tourism

An editorial on city tourism written in 2020 must discuss the impact of COVID. In many cities, an almost complete breakdown of economic income from tourists during at least some months this year has led to dramatic effects for the travel, hotel, gastronomy, entertainment and other related businesses. At the same time, residents whose income does not depend on tourism and the few remaining tourists appear to be relieved as that they can now enjoy cities without large crowds. “Enjoy Kyoto as it is supposed to be” or “the true Kyoto” have been trending during spring. This year proportionally more Kyoto residents could enjoy the cherry blossoms at major sights instead of travelling out of the city to avoid the crowds.

More broadly the current situation has been bringing more public awareness to the wider costs related to mass tourism. This relates to environmental pollution but also the strain it puts on the population. Therefore, a renewed discussion regarding an “optimal tourist number” seems appropriate. This should not only consider the local economy and interests of residents but also the tourists themselves. For tourists, with long queues and congestion, the negative experiences due to overcrowding might outweigh the gained satisfaction. Aiming to quantify this will be extremely difficult not least due to the aforementioned diversity of the tourists. Furthermore, there is the additional challenge to estimate trips not being made or diverted due to crowding. With more and more travel options becoming available, a public perception of a city being notoriously overcrowded might in the longer run lead to less visitors, or at least less high-spending visitors who are able to choose their destinations carefully. In terms of transport policy, this raises the question as to how much improving low-cost accessibility for mass tourism through new systems is really beneficial.

Returning to the immediate policies during this COVID crisis, measures to promote the resurgence of at least some safe level of tourism are controversial. Japan, being an island nation, has the advantage that domestic and foreign tourism can be more easily separated. Currently, with relatively low COVID case numbers in Japan, the Japanese government has launched the “go-to” campaign. Residents in Japan who want to travel to different prefectures can apply for discounted travel and hotel tickets. Participating travel agencies and hotels have a set contingent of discounted tickets or hotel rooms which can booked via online platforms. A further recent “go-to-eat” campaign is designed to encourage people to return to restaurants. Both campaigns are extremely popular. At the same time, the fairness, risk and usefulness of these schemes are critically discussed by the public. It is probably too early to judge whether these campaigns have been good overall. More generally, as a research question, it clearly highlights a conflict between risk and responsibility related to transport and economic policies that has been sharpened by COVID19. Promoting inter-city travel activities by raising accessibility (be it by pricing or reopening/ improving the transport services) will lead to more exposure of a resident population but contribute to national and local economic welfare. To control the risk, the aforementioned new data collection possibilities can play a key role. Predicting trends in tourist numbers and potential crowding near sights or in downtown areas has gained in importance as it determines both tourist satisfaction as well as risk exposure for tourists and the local population. Therefore, aiming for sensible (tourist) access restrictions to the city overall or to particular places based on mobile phone statistics and stationary sensor records appears to be another important and urgent research direction.

Concluding remarks

This editorial has argued that the estimation of city tourism flows is a research area deserving more attention not just from a tourist management point of view but from an overall, city level travel demand estimation point of view. There are fundamental differences with respect to demand dynamics and decision-making aspects between resident flows and visitor flows. These deserve more attention for transport planning in cities where tourists make up a substantial part of the car traffic, public transport demand and pedestrian traffic. Various new data sources have been noted that can support this task, but all have challenges to convert the information into flow estimates. Furthermore, how to define the “utility” of a particular tourist trip has been singled out as a difficult task. The COVID related breakdown of tourist activities and the aim to recover city tourism to some safe level has only increased the urgency of these research needs.

Acknowledgements

I would like to thank members of the project “Research into the development of a traffic flow estimation system for the purpose of understanding tourism flows” (Ministry of Land Infrastructure, Transport and Tourism of Japan “CART” scheme) as well as of the project “Tourists’ Flow Patterns Identification and Information Provision for Safe Evacuation” (Japan Science and Technology Corporation, grant number JPMJSC1805) for various discussions that have stimulated this editorial. In particular to mention are Nobuhiro Uno, Junji Nishida, Fumitaka Kurauchi, Toshiyuki Nakamura, as well as Yuval Hadas.