Spatial spillovers of tourism activity on housing markets: the case of Croatia

Several authors have stressed rising housing prices associated with intensifying tourism activity as a significant negative externality of tourism development (e.g., Mikuli, Vizek, Stoji, Payne, eh asni, & Barbi, 2021). The relationship between levels of tourism activity, typically measured in terms of arrivals and/or overnights, and housing prices has also been confirmed empirically in a number of studies (e.g., Biagi, Lambiri & Faggian, 2012; Biagi, Brandano & Caudill, 2016; Balli, Balli, Flint-Harle & Yang, 2019; Paramati & Roca, 2019, Churchill, Inekewe, & Ivanovski, 2021). This negative side-effect of tourism development gained new momentum with the proliferation of Airbnb and other peer-to-peer platforms that initiated a structural change in accommodation capacities at many destinations, essentially by turning flats into short-term rentals (STR) aimed at tourists (Dolicar, 2019). The negative consequences of these developments are well-documented in studies set within the context of “bucket-list” destinations struck by over-tourism, like, for example, Barcelona, Dubrovnik, New York, or Venice. Even the COVID-19 pandemic, which has hit the tourism sector very hard and led to a short period of under-tourism, did not stop this trend, with some European towns like, e.g., Split and Venice again being overrun by tourists during the peak of the 2021 summer season.Generally, besides rising housing prices, negative consequences associated with STR-induced overtourism are gentrified city centers (e.g., Wachsmut & Weisler, 2018; Ardura Urquiaga, Lorente-Riverola, & Ruiz Sanchez, 2020), crowdedness (e.g., Park & Agrusa, 2020), and increasing retail prices (e.g., Stynes, 1997, Gholipour, Tajaddini, & Andargoli, 2021), to name only some of the most significant ones. Together, these effects significantly contribute to a lowering of quality-of-life of residents in tourist destinations (Biagi, Ladu, Meleddu, & Royuela, 2020), especially for residents who do not own a property and/or have no significant benefits from the local tourism industry, either direct (e.g., via employment) or indirect ones (e.g., boosting economy). However, what if these effects were not constrained to neighborhoods in the area or destination under investigation (e.g., Zou, 2020) but rather extended to other regions via spatial spillovers? On the one hand, this would imply that the pressure of tourism activity on housing prices is an externality that also affects those who do not have any direct or indirect benefits from tourism activity, as described above. On the other hand, this would also imply that actions taken at the local level, like housing policies, aimed at providing relief to residents who have difficulties affording housing at popular tourist destinations, should, in fact, be developed more holistically by taking into account tourism impacts on real estate markets at a regional or supra-regional, rather than focusing only on the local level.As described above, potential regional spillovers have not been addressed in the tourism, housing, nor regional economics literature so far, which is a research gap this study intends to fill. The empirical analysis is set within Croatia, an increasingly popular Mediterranean destination, ranked 27th according to the World Economic Forum’s Travel & Tourism Competitiveness Report, putting it at a similar level like, e.g., Greece (#25) (WEF, 2019). Tourism activity in this country is densely concentrated in proximity to the sea and along the whole coast, whereas there are only a few tourist spots in the continental parts of this country. This is also reflected in a share of 97.3% of all accommodation establishments located in one of two of Croatia’s NUTS-2 regions, i.e., Adriatic Croatia (as opposed to Continental Croatia), which is also ranked first among all European NUTS-2 regions according to accommodation capacity (Eurostat, 2021a). At the same time, Croatia also has the largest share of private accommodation in overall tourist capacities compared to its Mediterranean peers. For example, the percentage of holiday and other STRs in total bedplaces in pre-pandemic 2019 was 61.3% in Croatia, whereas the same share was 34.5% in Italy, 32,2% in Greece, or 24.6% in Spain (Eurostat, 2021b). Accordingly, although geographically small, Croatia is ideal for examining regional spillovers as described above because its tourism sector heavily relies on apartment houses and is highly concentrated in its coastal area.The research question we are addressing in this paper should also be placed in the context of the ever-increasing global tourism demand. International tourism receipts increased 4.9 percent in real terms to reach US$ 1.34 trillion in 2017, whereas tourist arrivals amounted to 1.336 billion in the same year (UNWTO, 2018). In addition, international tourist arrivals worldwide are expected to increase by 3.3% annually to reach 1.8 billion by 2030 (UNWTO, 2017), which suggests popular destinations already overwhelmed with tourists will start to experience more negative externalities of tourism activity, including the degradation of sociocultural and environmental conditions and the rise of house prices and rents coupled with declining housing affordability. In geographically smaller or island countries which depend more on tourism receipts, the emergence of over-tourism could potentially be more damaging to the local housing markets due to limited availability of building plots, higher population density, and often rigid urban zoning rules. In such countries, house price hikes taking place in a very popular tourist destination (local government units - LGUs) could easily spillover to neighbouring LGUs, then regions and eventually the entire country. For this reason we study local/regional house price spillovers which are due to tourism activity in Croatian LGUs.In order to understand the spillover process better, we will use spatial panel data estimators to model house prices and its determinants, which in turn will enable us to detect the main spillover characteristics and modalities. To the best of our knowledge, this would be the first study of its kind in the literature. The only other study addressing this issue is Kavarnou and Nanda (2018) who use dummies for neighboring regions as controls for spatial spillover effects, which is a static and rather limited way of addressing this issue.The empirical part of our analysis relies on techniques from the family of spatial panel estimators. We use the Durbin spatial panel autoregression technique (DSM) (LeSage and Pace, 2009; Elhorst, 2013). This technique enables us to establish a direct relationship between the dependent variable and its effects in other spatial units (cities or regions), spatial effects of independent variables such as measures of tourism activity and unobserved spatially correlated heterogeneity. Moreover, we employ different types of spatially weighted matrices, symmetric matrices that define relationships between units in space. To this end, we explore whether spatial effects are limited to neighboring municipalities and cities or do they exert wider scale. Another advantage of spatial econometric analysis is the ability to estimate direct and indirect effects of the observed process. The change in individual city or region generates two types of impacts. A direct impact on itself and an indirect impact that goes first to other spatial units and partially returns through feedback loops (LeSage and Pace, 2009). These feedback loops arise because each spatial unit is considered a neighbor to its neighbors so the impact passing through neighboring units will create a feedback impact on the initial unit itself. There is reason to expect such feedback loops will arise with changes in tourism activity and dynamics of the housing market. To the best of our knowledge, such analysis has not been performed in context of tourism sector or its relationship with the housing market in general. Our dataset combines variables constructed from several reliable data sources. Most of the variables, including the tourism related measures, come from the National Statistical Office of Croatia, the official focal point for statistical data collection. We use five different proxies for tourism activity and intensity: number of nights spend per inhabitant, number of arrivals per inhabitant, the share of private accommodation in total tourism accommodation, the share of rental housing in total housing stock and the length of stay of tourists.This database is supplemented with datasets obtained from the Ministry of Finance, i.e. its Tax Office from which Property income and Average wage data come from. Finally, the Institute of Economics Zagreb (EIZ) provides the data on housing transactions and housing and construction land prices from their annual reports on real estate trends prepared for the Croatian Ministry of Construction. The analysis covers the 2012-2019 period and contains 556 Croatian cities and municipalities in which real estate transactions took place over the years analyzed.