Interconnectedness of international tourism demand in Europe: A cross-quantilogram network approach

Highlights

• We study directional interconnectedness in quantiles of tourism demands in Europe.

• Networks are dense when tourism demand growth is small.

• Networks are sparse when tourism demand growth is larger.

• Most of the directional dependence is capture in first three months.

• Using ERGM, we identify reciprocity as a common feature of tourism demand networks.

Abstract

We study the interconnectedness of international tourism demand changes among 30 European countries. Using cross-quantilogram analysis, we estimate the strength of the directional (lead/lag) relationships of the international tourism demand of European countries in percentiles (10th, 50th, 90th). The complex interconnectedness of international tourism demand is studied within networks, where a fixed number of vertices represent countries, and oriented edges represent the presence of a directional relationship between the international tourism demand of two countries. A comparison of these networks reveals the following regularities. First, we find obvious asymmetry across percentiles, where demand behaves much more similarly during times of crisis (10th percentile) compared to tranquil periods (50th percentile). The interconnectedness of these networks almost diminishes when the international demand for tourism increases sharply (90th percentile). Second, we observe that the interconnectedness does not change much among the short- (within 3 months), mid- (up to 6 months) and long-term (up to 9 months) lead/lag relationships, which leads us to conclude that much of the interconnectedness of international tourism demand is driven by dependence during the first three months. On the basis of these findings, we review the possible forces that may drive the formation of the resulting complex structures using exponential random graph models. Our third finding is that there is a tendency for the relationships of the international tourism demand among the various countries to be bidirectional. Finally, our fourth new finding is that the interconnectedness of markets during sharp declines in tourism demand tends to increase for Central and Eastern European (CEE) countries, and those that are less developed in terms of their relative sector size to the size of the economy.

Introduction

Since the early work of Barabási and Albert [1], networks have been found, created and studied in many socioeconomic areas. As is documented in the recent review of Martí et al. [2], in the past few years, networks have been particularly used in finance [3], [4], [5], business and trade [6], [7], [8], [9], [10], supply chain networks [11], [12], [13], transportation [14], [15], [16], [17], [18], [19], social relationships [20], [21], [22] and also in tourism [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39].

In this study, we contribute to the latter strand of the literature. The tourism sector includes many agents such as visitors; accommodation providers; food and beverage services; recreation and entertainment; transportation companies; and travel services, such as travel agencies, travel guides, tour operators and insurance companies, and it is therefore believed that the tourism sector consists of a complex system of relationships [23]. Motivated by this idea, in this paper, we study how the international demand for tourism is related among 30 European countries. We take a macro-level perspective, where countries represent tourism destinations, and international tourist arrivals at a given country represent the international demand for tourism services in a given country. Therefore, among the many agents that operate in the tourism sector, we specifically consider international visitors. At the macro-level, we aggregate international demand across the whole country. We acknowledge that each country can influence local tourism sectors through tourism policies, however, it is likely that these, in turn, are manifested into the overall level of international tourism demand for the whole country.

We create a real-world network, where nodes represent countries, and edges are created according to the dependence measured between the numbers of arrivals of non-residents (demand) to a given country. Such complex networks are of interest to a wide range of agents. For example, a dense network suggests that the international demand of different countries is interrelated, and tourism policies, at a country level, could be collaborative (for an example of cross-country regional collaboration, see [40], [41]). Out- and in-degrees can be interpreted as measures of relatedness of among the tourism policies of countries. Put differently, in-degrees measure the sensitivity of international tourism demand to the development of international tourism in foreign countries. On the other hand, out-degrees indicate whether the international tourism demand of a given country could be used to forecast development of international tourism demand in foreign countries (for example, by including past observations of tourism demand into the forecasting equation of tourism development of foreign countries).

Real-world networks have already been used to describe the tourism sector [42]. The websites of tourism destinations on Elba Island showed little interconnectedness in a study conducted by de Fontoura and Baggio [26]. Hernandez and González-Martel [27] modeled a supply network and the interrelationship among lodging, services and tourist behavior in the tourism destination of Gran Canaria Island. There was a positive correlation among the degree of substitutability, closeness centrality, and betweenness centrality for tourism flows in China in a study conducted by Hong and Tao [28]. The binary and value network of tour operators and travel agencies, which are entities of a distribution channel, was illustrated by Tran et al. [22] in a case study of Hanoi, Vietnam. Networks can be used to help policy makers and tourism service providers improve their planning of a given destination development [29], [30], [31], [32], [33], [34], [35], [39].

Our study is most closely related to one conducted by Miguéns and Mendes [36], who created an oriented network of international tourism arrivals in 2004, with 208 nodes representing both countries and territories. Oriented edges were created if tourists traveled from node $i$ to node $j$. The resulting network led to 5775 weights, a dense topological structure with small shortest paths and exponentially decaying in- and out-degrees. The weights followed a scale-free power-law distribution. Provenzano et al. [43] created and studied a network of bilateral tourism flows within European countries using data from the United Nations World Tourism Organization (UNWTO) over the period from 1995 to 2012. They reveal that the network’s density increased, the average path length decreased, while also tourism evolved towards higher concentration in a lower number of areas. Our approach differs in several ways. First, compared to Miguéns and Mendes [36] we use a more homogeneous sample that includes 30 European countries, and each country represents a node in the network. Our sample selection is motivated by the fact that these countries are in the same region, and almost all of the countries in our sample are members of the European Union and subject to similar regulations at the European level. Second, an oriented edge is created if Granger causality in quantiles is confirmed from international tourism demand of country $i$ to country $j$. Third, using monthly data from January 2004 to May 2018, we study the determinants of international tourism demand.

The remainder of this paper is structured as follows. Section 2 presents the data and describes the methodology. In Section 3, we discuss the results, and Section 4 concludes.