Every journey creates connections between destinations, gradually forming a vast global network that remains largely invisible to us. Mapping those connections can reveal key insights, for example, how tourists move around the world and the very different roles destinations play within their journeys.
“We’ve landed in Rome and spent a few days sightseeing. Now where are we off to?” This is a classic scenario for millions of visitors to Italy every year. Indeed, the ‘next destination’ question is essential to hospitality operators around the world.
A recent study by Yong Chen and Giuliano Bianchi from EHL Hospitality Business School attempts to answer this question and has provided, for the first time, a glimpse of what the global destination network looks like and why it matters for academics, industry and the general public.
The global destination network was mapped using hotel reservation data published by Booking.com in 2021 and is the largest and most complete data set of tourists’ trips to date at the global level. The data set includes a total of 217,686 distinct trips taken by 200,153 travelers, departing from five countries, across 39,901 cities in 195 countries over the course of one year. It spans from December 31, 2015 to February 28, 2017.
A tourist destination network is formed through individuals traveling from one destination to another during their trip, whereby geographic locations are connected with each other. The resulting connections between destinations have no bearing on interpersonal relationships between tourists but reflect the accessibility of these locations in various dimensions. It stands to reason that a tourist landing in Milan would then visit Florence, Rome and further south to Naples or, for the more intrepid traveler, even Palermo in Sicily. A host of factors influence these movements, including tourists’ economic, physiological and psychological characteristics as well as broader geographic, economic, social, cultural, political and technological factors.
The 2 Trillion-Dollar Question
The global tourist destination network is one of the largest and most important social and economic networks. It encompasses over 1.5 billion annual visits by international tourists, generating $1.9 trillion in expenditure (UN Tourism, 2026). Our study was more limited, covering hotel reservation data from five countries of origin. It is, however, highly representative in the sense that all 195 countries in the world are included.
The global destination network shows some key properties that are not found in other social and economic networks, nor do they resemble the three canonical network models, which are believed to account for a large proportion of network phenomena in the real world.
High Connectedness of Global Cities Through Tourism
None of the five destination networks are fully connected. Each contains a component, i.e. a subset of cities that are connected with each other but disconnected from all other cities in the entire network. The existence of components suggests that tourists are “trapped” in “local” networks and cannot “escape” for various reasons. A typical, and also an extreme, case is bilateral tourism between two cities. In bilateral tourism, there exists the smallest component that contains only two cities, which are each other’s destination. Even if bilateral tourism is predominant in the global market, as the network becomes large and complex, individual city pairs are less likely to be independent of and disconnected from each other. This will eventually give rise to the largest component(s) which contain the lion’s share of cities.
The existence of the largest component(s) is one of the key features of the global destination networks that are built up on a large number of individual trips. As we show, the largest components account for between 96.0%, or 6,720 cities, in the Bartovia network and 99.8%, or 10,625 cities, in the Devil Fire Empire network. This suggests that the largest components can represent the entire destination networks, and the destination networks are highly connected.
It is worth noting that destination networks are directed by nature, meaning that a trip from city i to city j, in which i is connected to j (i.e., j is the destination), is not the same as a trip from j to i, in which j is connected to i (i.e., i is the destination), and that the occurrence of a trip from i to j, or lack thereof, does not imply that there must be a trip from j to i, or lack thereof, and vice versa. A subset of cities in which a tourist departs from any city can reach, through directed path(s), any other city in the set is referred to as the strongly connected component of a network. The smallest strongly connected component is simply an isolated city pair in bilateral tourism. Of interest to us is the largest strongly connected component (LSCC) of a network, because it contains the maximum number of cities, each of which can be reached by tourists departing from any city.
To the extent that the strongly connected component accounts for a substantial proportion of a network, it can simply represent the entire network in terms of both size and directionality. The LSCCs of the five destination networks account for between 76.5%, or 3,101 cities, in the Techerkistan network, and 96.0%, or 6,719 cities, in the Bartovia network. In the Techerkistan network, for instance, any tourist originating from Techerkistan, regardless of their first destination, can reach each of the remaining 3,100 cities in sequence without having their trip interrupted. This does not only mean the vast majority of the cities being connected by tourists’ trips but also suggests the abundant supply of transportation and accommodation in these cities that makes tourists’ trips physically possible and economically feasible.
Scale-Free Properties and Hub Cities
The destination networks show a robust scale-free property, meaning that the number of cities in the networks decreases sharply as their connections increase. This results in a sufficiently large proportion of cities that are connected with only a few other cities while a small proportion of cities that are connected with a sufficiently large number of other cities. In the five destination networks, each city is connected on average with between 4.4 cities in the Bartovia network and 10.2 cities in the Gondal network, while the maximum connections are between 176 cities in the Bartovia network and 708 cities in the Gondal network.
The scale-free property suggests the existence of hub cities that are connected with a large number of cities in the networks. Not only are these hub cities themselves destinations as in bilateral tourism but they are also gateways to other cities. This does not necessarily mean that hub cities must be geographically central, but they are “central” in terms of their accessibility to a large number of other cities. A hub or gateway city in a destination network is accessible to all other cities with only one or a few connections, simply because it is connected with a large number of other cities. Hence, average path length, i.e., the number of connections, between a hub city and other cities is much shorter than those between peripheral cities. Yet it is not possible to pinpoint what these hub cities are because all cities are anonymized in the data set.
Scale-free properties have been widely documented in many networks, including the Internet, the Web, citation networks, as well as biological networks, which has led researchers to believe that scale-free networks are universal. We provide further evidence in this regard.
The “Big-World” Phenomenon
The destination networks have exceedingly long diameters, which are defined as the longest shortest paths between any two cities that are connected in a network. In a connected destination network, if one picks two cities i and j at random, the shortest path between the two is the diameter of the network. The phenomenon of “six degrees of separation” or “small-world” simply suggests that social networks have short diameters. Even if a network is completely random, i.e., any two nodes are connected uniformly with some probability, as the network grows, its diameter will increase logarithmically with its size, giving rise to the “small-world” phenomenon.
However, the destination networks have much longer diameters than all other social networks, ranging from 14 in the Techerkistan and Devil Fire Empire networks to 25 in the Elbonia network. This means that any tourist departing from their first city, regardless of what it is, needs to make at least 14 to 25 stops before reaching a randomly chosen city in the network. The simulations also show that the destination networks have diameters that are two to three times those of the simulated networks. All evidence thus suggests that the destination networks exhibit a “big-world” phenomenon, in contrast not only to the “small-world” phenomenon in social networks but to the canonical network models. This implies that the destination networks need to contain a nontrivial number of peripheral cities that are farther away from hub cities to elongate the diameters.
Seeking ‘Off-the-Beaten-Path’ Destinations
The “small-world” phenomenon emerges in social networks because the connections are built up on interpersonal relationships whose value is predicated on how “close” these relationships are. Hence, minimizing the number of connections between people is essential for many social networks, such as online social networks and friendship networks, for it helps accelerate information diffusion and builds up trust and closeness. However, tourists do not seek to minimize the number of connections between cities. Quite the contrary! Many expect to visit a sequence of cities that are geographically and/or culturally distinct so as to enhance the novelty and excitement of their overall travel experience. The search for novelty, rather than familiarity, leads tourists to favor peripheral cities that are not directly connected to hub cities and are therefore less visited. It is the inclusion of peripheral cities that extends the diameters of the destination networks.
Why the Global Destination Network Matters
When tourists land in Paris, there is a fair chance they will visit other French or European cities. But which ones? And in what order? The answers to these questions hold the key to billions of dollars in revenue for hospitality operators.
Our study shows that this network is simultaneously highly connected, concentrated around powerful hubs and surprisingly extensive at its edges. That combination distinguishes global tourism from many of the social networks we already understand.
Indeed, where a destination sits within a wider journey (i.e., the web it forms) matters! Understanding the connections between destinations can ultimately provide a richer picture of tourism flows and help hospitality businesses and destinations think beyond individual stays towards the broader journeys travelers make.
