Territorial dynamics of tourism in Romania: a long-term perspective (1990-2016)

The potential role of tourism in enhancing a balanced territorial development is officially acknowledged by the EU. However, little is known regarding the extent to which tourism in Romania contributes or could contribute to reducing regional inequalities. The present study approaches the matter of the growth of tourism demand and supply in relation to the type of territory (lagging, intermediary, leading), at both regional and local levels. The results indicate a significant geographical concentration of tourism activity, despite a noticeable diffusion of tourism across the country in the last decades. This derives from the fact that the leading territories, which already took the lion’s share of the tourism supply and demand also retain the highest part of the increases in tourism demand. As a consequence, tourism sector stands out more as a factor that can increase territorial disparities than contribute to their reduction. However, a considerable potential for tourism growth in terms of tourism supply characterizes the lagging regions. Accordingly, differentiated tourism public policies related to the type of territory could be considered.


Introduction
The subject of tourism' spatial distribution and its territorial dynamics has been largely approached in the scientific literature (Saarinen et al., 2017;Sarrion-Gavilan et al., 2015;Yang & Wong, 2013;Luo & Yang, 2013). It gained a certain appeal for policy-makers, given the fact that the level of concentration/dispersion of tourism activities is directly related to the impact those activities have or may have on the territory (Cornelissen, 2005). At the same time, there are still debates whether tourism is a means of ensuring a balanced development of the territories or, on the contrary, a determinant of regional inequalities (Andraz et al., 2015;Tosun et al., 2003). The presence of the tourism sector in development strategies, as a 'mechanism of regional development' (Chhetri et al., 2013) is more and more pronounced and, therefore, it is essential to understand how tourism evolves spatially and temporally, in order to maximise its benefits over a territory.
Tourism activities, by their nature, are susceptible to spatial concentration, since tourism investors locate their businesses in the proximity of tourist attractions (Majewska, 2015). It is thus not surprising that the tendency of tourism to concentrate spatially has been widely studied, with a particular focus on its relevance for tourism policy-making (Stankov et al., 2017;Lau & Koo, 2016;Sarrion-Gavilan et al., 2015;Kang et al., 2014;Chhetri et al., 2013;Zhang et al., 2011). The European Commission (2010) acknowledged the growing economic importance of tourism and highlighted its potential for economic development and social integration, especially on rural, mountainous, and peripheral areas. The transversal character of tourism activities calls for reliable connections between tourism related strategies and policies from other fields, such as transportation, competitiveness, employment, or environment. This relationship owns its origins to the fact that in many contexts, tourism is perceived and approached as an economic solution when other resources are missing or when main economic activities tend to decline in importance (Boujrouf et al., 1998;Ibănescu, 2015). The spatial dimension is of central importance in tourismrelated policies since the nature and potentialities of a territory influence the evolution of tourism activities within and represent essential factors in establishing development priorities (Iatu & Bulai, 2011;Saarinen et al., 2017).
Romania suffers from significant territorial development imbalances at several geographical scales, even since its formation as a nation-state (Muntele, 1998;Ungureanu, 2005). These long-date imbalances can be observed between historical regions (Groza, 1998), counties (Benedek & Török, 2014), or even between urban and rural areas, where they are probably among the most prominent (Sandu, 2011). Measures intended at increasing territorial cohesion were adopted during the interwar and communist periods (orientated especially towards education, infrastructure, and various economic activities), although with limited results (Kurko, 2010;Popescu, 1994). Following the fall of the communist regime in 1989, territorial disparities started to significantly increase (Istrate & Horea-Șerban, 2016;Zaman et al., 2013,). During recent decades, some territories adapted better to the new economic realities and gained significant advantages, while other territories struggled to keep the pace. Within this challenging context, tourism has recently been discussed as a possible means of reducing territorial development imbalances, especially between the central/developed regions and the peripheral ones (Roberts & Hall, 2001;Ibănescu et al., 2018). Understanding the past evolution and the present state of tourism in a territory is essential for determining the potential future trends, impacts and the best ways to respond (Kang et al., 2014).
The purpose of this study is to identify and discuss the main tendencies in the territorial dynamics of tourism in Romania over the last decades. This kind of analysis is necessary in order to see if tourism in Romania may act as a channel for convergent, sustainable, and balanced territorial development. To our knowledge, a study of this kind has not been conducted yet for the Romanian case. The moment of the country's integration in the EU is regarded as a milestone in this study. Comparing territorial dynamics before and after this moment allows us to see the degree to which tourism evolution in Romania is connected to the European Commission's views on the potential role of tourism in enhancing a balanced territorial development.
The first part of the study consists in a literature review on the main methodological approaches on tourism' territorial dynamics. The data and methods employed are presented in the second section of the study. The results are structured in three subparts: (1) an overview of tourism's evolution before the fall of the communist period, (2) an analysis of the tendencies of dispersion or /concentration of tourism supply for the period 1990-2016 and (3) a comparison between different types of territories based on the evolution of tourism demand and tourism supply. Lastly, discussions and recommendations are drawn based on the study's results.

Methodological approaches in determining tourism territorial dynamics: a literature review
The territorial dynamics of tourism represent a subject of interest for researchers from multiple fields. Their utility and impact were discussed mainly with reference to tourism policy-making (Stankov et al., 2017;Majewska, 2015;Sarrion-Gavilan et al., 2015;Chhetri et al., 2013), but also in relation with the business sector interests (Luo & Yang, 2013;Chhetri et al., 2013). A high attention was given to the territorial development induced by tourism activities (Schirpke et al., 2017;Salvatore et al., 2018), especially to the role played by tourism on regional disparities (Majewska, 2015;Goh et al. 2014;Yang & Wong, 2013;Ivy & Copp, 1999). Consequently, various approaches and diverse methods have been developed in order to analyse the spatial dimensions and impacts of tourism in different territories (Majewska, 2015;Zhang et al., 2011). Following a literature review, the authors concluded that the existing quantitative methodological approaches vary significantly in terms of complexity, purpose, and techniques employed. However, they can be largely classified into four main categories.
A first category consists of approaches that provide an image of the spatial and chronological distribution of tourism activities through simple methods and easyto-communicate visual results (Rogerson, 2013;Van Doren & Gutske, 1982). Although these studies lack the depth provided by a more complex methodology, the simple mapping of tourism variables represents a solid starting point concerning the understanding of the tourism' territorial dynamics' implications. The choropleth mapping of different tourism related variables provides a proper overview of the characteristics of tourism and its tendency of concentration or dispersion at a particular moment or over a period of time. This approach has been successfully used for analysing accommodation capacity (Ferreira & Boshoff, 2014;Rogerson, 2013), tourism demand (Iliev, 2018;Iyv & Copp, 1999), tourism receipts (Van Doren & Gutske, 1982), or tourism investments (Cornelissen, 2005).
A second category of approaches incorporates various GIS-based analysis techniques that the determine territorial evolution of tourism, such as computing spatial descriptive statistics of tourism indicators for various types of territories (Sarrion-Gavilan et al., 2015;Rogerson, 2013), or modelling (changes in) catchment areas for various destinations (Ibănescu et al., 2016;. A third category of approaches emerged during the last decades by incorporating inferential Exploratory Spatial Data Analysis (ESDA) in tourism studies. ESDA is defined as 'the collection of techniques to describe and visualise spatial distributions, identify atypical locations (spatial outliers), discover patterns of spatial association (spatial clusters), and suggest different spatial regimes and other forms of spatial instability or spatial non-stationarity' (Anselin, 1998). Applying ESDA in tourism research generally implies computing global and local spatial autocorrelation indexes, which allows the analysis of spatial concentration tendencies (Stankov et al., 2017;Yang & Wong, 2013;Luo & Yang, 2013;Zhang et al., 2011). Global indexes detect spatial tendencies 'from the general perspective by incorporating all samples' while 'the local ones only focuses on the specific relationship in a particular location over space' (Yang & Wong, 2013).
Researches employing ESDA in the analysis of spatial dynamics of tourism usually follow two main approaches. A first approach is based on providing a diachronic analysis concerning the concentration or dispersion tendencies of tourism activities in a particular territory (Stankov et al., 2017;Sarrion-Gavilan et al., 2015;Yang & Wong, 2013;Zhang et al., 2011). For example, banking on this method, Yang & Wong (2013) identified significant tendencies of concentration for inbound and domestic tourism flows in China during the period 1999-2006 and investigated the tourism hot-spot areas for both types of tourists through local indicators of spatial autocorrelation. A second and more complex approach is switching from uni to bivariate ESDA, which allows for detecting the potential factors for the spatial patterns identified (Romao & Saito, 2017;Majewska, 2015, Luo & Yang, 2013. For example, Luo & Yang (2013) employ ESDA measures for identifying and explaining the spatial patterns of hotel located in Chinese cities, by making use of both univariate and bivariate Moran Statistics. While the univariate global and local indicators allowed for the identification of hotels spatial concentration over time, the bivariate statistics provided further information by giving insights into the relationships between the geographical patterns identified and variables that concern the economic context, foreign investment dependence and local tourism development, which are hypothesized as potential factors of the existent spatial distribution of tourism activities. Majewska (2015) added spatially weighted location quotient, Herfindahl index, and tree-clustering analysis to the results obtained through spatial auto-correlation in order to address the issue of the relation with the neighbourhood. Romao & Saito (2017) employed a regression model in order to provide an evaluation of the relations between the spatial patterns of tourism and a series of variables related to the tourism industry and economic development in Japan. A fourth category of methodological approaches incorporates methods aimed at studying tourism regional spillovers. By approaching the spillover effects, some researchers highlighted the benefits some regions got through the development of tourism activities in the adjacent regions (Majewska, 2015;Yang & Wong, 2013;Zhang et al., 2011). Those results are considered as being of particular interest in the elaboration of tourism strategies focused on regional cooperation. Related to the abovementioned results, recent studies identified the opportunities and necessities from a geographical point of view for investments and budget allocation, in terms of infrastructure or tourism products' creation (Sarrion-Gavilan et al., 2015;Yang & Wong, 2013).
This research investigates the main tendencies in the territorial evolution of tourism in Romania over the last decades, by employing methods from the first three categories. The authors hypothesize that the spatial distribution of tourism in Romania manifested a tendency of dispersion during the last decades, which set a favourable context for tourism to play a significant role in reducing development inequalities between leading, intermediary, and lagging regions. The study also inquires the particularities of tourism evolution with reference to various types of territories.

Data and methods
Two approaches have been developed in order to test the hypothesis and attain the objectives stated above: 1) an exploratory spatial data analysis in order to test for the hypothesis stating that Romanian tourism supply and demand manifest a tendency of dispersion during the last decades, and 2) a GIS-based spatial analysis approach to find out which type of territories (leading, intermediary, lagging) are taking the lion's share of Romanian tourism growth. This second approach makes it possible to see whether tourism could be regarded as a factor of reducing territorial economic imbalances or, on the contrary, as a factor contributing to their increase. The exploratory spatial data analysis (ESDA) has been conducted by comput-ing spatial auto-correlation indexes. Spatial auto-correlation indexes allow researchers to answer a fundamental question in geography: 'Is the spatial pattern displayed by the phenomenon significant in some sense and therefore worth interpreting?' (Getis, 2007). In our case, this translates into deciding whether the territorial pattern displayed by tourism supply and demand is significant in some sense and therefore worth interpreting. For the purpose of the present research, the authors computed Moran's I global spatial autocorrelation index for 2946 LAUS2 in Romania, for each year between 1990 and 2016 in order to have a glimpse of the tourism dynamics over the Romanian territory. Source: own work * One important issue concerning spatial data sets deals with the changing number of LAU2s during the last 25 years. Romania has passed from 2,948 local administrative units in 1991 to 2,946 in 1994 and 3,181 in 2016 (Bucharest counted as one single unit including all its 6 sectors, which are otherwise administrated separately as distinct LAU2s). However, changing the number of observations over time could alter the inferential statistical analysis. Thus, a single spatial data-set was created and employed for the whole range of the analysis. Tourism supply and demand data were thus re-aggregated for matching the chosen data-set for every single year of the period 1990-2017. The only data-set that could allow the aggregation of the data provided by the NSI is the one from 1994 comprising 1,946 administrative units. Thus, this was chosen to conduct our analysis at the local level. ** Data sources: NSI (2018) *** NUTS2 = basic regions for the application of regional policies in the 28 EU Member States (corresponding to the Romanian Development Regions), NUTS3 = small regions for specific diagnoses (corresponding to Romanian counties), LAU2= municipalities or equivalent units in the 28 EU Member States (corresponding to communes, towns and cities in Romania). Table 1 shows tourism indicators that have been included in the analysis (both in the first phase of the exploratory spatial data analysis and in the second phase of the GIS-based spatial analysis). Tourism territorial dynamics have thus been analysed from two distinct perspectives: demand and supply. Both supply and demand have been quantified by employing official data published by the Romanian National Institute of Statistics.
The GIS-based spatial analysis approach was chosen in order to find out which types of territories are taking the lion's share of Romanian tourism growth. Romanian administrative units had to be classified into different types according to their economic status (leading, intermediary, lagging), but also according to their tourism resources supposed to generate different types of tourism (spa tourism, Seaside & Danube, mountain, urban, rural). Thus, if peripheral and intermediary regions take the lion's share of tourism growth one could argue that tourism is reducing territorial inequalities, or at least has the potential to do so. On the other hand, analysing tourism's evolution by the type of tourism resources is not answering directly the main research question. However, it complements the findings by showing which types of tourism resources generate the most dynamic trends.
The Romanian territory has been classified into leading, intermediary and lagging at three distinct geographical scales (NUTS2, NUTS3 and LAU2). Romanian regions (NUTS2) and counties (NUTS3) have been classified into leading, intermediary and lagging depending on their GDP/inh. at the beginning of the analysis period (2000). Given the high discrepancies in terms of GDP/inh. between the NUTS 2 Bucharest-Ilfov and the next NUTS 2, the creation of a distinct category for the capital region seemed justified. Tables 2 and 3 show the number of regions included in each category and the value of GDP/inh. for each of the four categories. Romanian communes, cities, and towns (all LAU2s) have also been classified into three categories, depending on their population densities. High population density is often regarded as a proxy for centrality as it generally occurs in cities or large rural communes, which act as local centres of development. On the other hand, sparsely populated areas are per se considered peripheral. Considering that the population density for the entire country was of 91 inh/km 2 in 2002, the following thresholds have been employed: above 120 inh/km 2 (high and very high densities), 60-120 inh/km 2 (medium or close to medium density) and below 60 inh/km 2 (low and very low densities). Table 4 shows the main attributes for each of the three categories of LAU2s. Finally, five types of territories according to their tourism resources have been defined: spa tourism, seaside & Danube Delta related tourism, mountain tourism, urban tourism and rural tourism. Table 5 shows the criteria and the number of LAU2s falling inside each of the five categories.

Tourism supply territorial patterns inherited after the fall of the communist regime
At the beginning of the post-communist period, the country's tourism territorial pattern was mainly characterized by a high concentration of tourism activities. The high territorial concentration of tourism activities is a characteristic of every nation in its first stages of tourism development. However, in the Romanian case, it is also a consequence of the tourism policies implemented during the communist period. The regime did not approach tourism as a priority and it mainly regarded it with reference to its recovery potential for the Romanian tourists, with the purpose of (re)consolidating their work capacity (Rădulescu & Stănculescu, 2012).    Source: own work *One LAU2 may be part of two or more categories as more types of tourism can take place inside the same LAU2.
The territorial dynamics of tourism during the communist period was mainly determined by two factors: (1) the communist administration's own criteria of localisation the tourism infrastructure and (2) a particular interest towards the development of mass tourism. The first factor refers to the prioritisation of investments in tourism in two categories of territories: (1) the most important cities from an administrative, industrial, and political perspective, which in many cases were lacking in tourism value and (2) areas in possession of significant resources for spa tourism. Thus, it is not surprising that in 1990 tourism activities were highly concentrated, with destinations located at the Black Sea summing up to over 73% of accommodation capacity in Romania (Dumbrăveanu, 2001).
The first decade after the Romanian Revolution of 1989 was characterized by a delayed development of tourism, which significantly influenced its territorial dynamics. On the tourism supply side, the main causes of the delayed development of tourism reside in the long process of privatization (Iațu, 2009), which was not effectively finished before 2000. The slow progress of the privatization process can be explained through a series of factors: (1) the complexity of the privatization procedures alongside with the absence of precise laws and conditions, that were universally applicable; (2) the low quality and standards of the accommodation infrastructure, inherited from the communist period; (3) the decline of the tourism demand, that discouraged potential investors. Two particular negative effects are the deterioration of the accommodation infrastructure, which was mainly determined by the lack of investment (Light, 2006), and the decrease of the accommodation capacity in the main tourism destinations of Romania, which triggered a less clustered pattern.
On the tourism demand side there was a significant decrease in the number of tourist arrivals, both in the number of domestic and in that of foreign tourists. The decline of domestic tourism was determined by the economic restructuring and a decline in the living standards, which determined tourism to be inaccessible to a high percentage of the population. On the other hand, the decline of inbound tourism had two main causes that reinforce each other. The first one resides in the absence of strategies aimed at raising the country's attractiveness for foreign tourists. The second cause is the low quality of tourism infrastructure and services that created a negative image of Romanian tourism abroad. These factors, associated with the necessity of increasing the prices for the accommodation services, in order to cover the maintenance costs and to pay employees, set the conditions for a continuous degradation of the accommodation industry during the 1990s. However, these tendencies have fundamentally changed from the years 2000s.

Towards a less clustered territorial pattern of tourism supply during the last 26 years
Looking at the changes in the number of LAU2s offering tourism accommodation services is one way of highlighting territorial and temporal patterns of tourism. Figure 1 highlights the existence of an expansion phenomenon as the number of LAU2s offering tourism accommodation has increased from 490 in 1990 to 948 in 2016 (that is from 16.6% of the total number of LAU2s in 1990 to 29.8% of the total number of LAU2s in 2016). Figure 1 also shows that most of the growth happened after the year 2000. The growth was boosted by a favourable economic context and by two changes that occurred in the hospitality industry: 1) the increasing number of private initiatives after a decade of failed attempts of privatization, and 2) the emergence of business tourism in cities and rural tourism in peripheral areas. These fuelled the development of the tourism industry and triggered the diversification of tourism accommodation supply and its territorial expansion. Nowadays, almost 1 out of 3 LAU2s is offering tourism accommodation services.
However, territorial expansion does not necessarily mean territorial dispersion. Tourism territorial expansion occurs when tourism is conquering new territories that previously registered no tourism activity. On the other hand, tourism territorial dispersion occurs when tourism demand/supply become less spatially clustered.

Source: own work
Tourism territorial expansion is evident in Romania, while tourism territorial dispersion may not be a reality, as positive evolutions of tourism activities in peripheral areas have been accompanied by an even more pronounced positive evolution of the main cities and tourism regions, thus questioning the existence of a genuine dispersion pattern. Therefore, one can reasonably ask if the Romanian tourism supply is either clustering or becoming dispersed. Inferential statistics in the form of global indexes of spatial auto-correlation are particularly useful for testing such a hypothesis. Figure 2 shows that the spatial concentration of the Romanian tourism supply decreased to a noticeable extent in the last 26 years, thus becoming less clustered. However, one cannot yet assume the existence of a dispersed pattern of tourism as the values of Moran's I are still very high in 2016. The seaside area, the most important cities, such as Bucharest, Sibiu, Cluj-Napoca or Brașov and certain mountain areas, as is the case with Prahova Valley, continue to concentrate most of the tourism accommodation capacity. For example, Bucharest has passed from 10,000-beds capacity in 1990 to 20,000 in 2016, with localities in the vicinity also significantly increasing their tourism accommodation capacity. In the meantime, 553 rural LAU2s (all LAU2s under 30 inh/km 2 ) had a similar absolute increase, from almost 16,700-beds capacity in 1990 to 27,700-beds capacity in 2016. Thus, the increase in the accommodation capacity of one single city equals the increase of the accommodation capacity of all the 553 LAU2s that have less than 30 inh/km 2 . Romania: a long-term perspective (1990Romania: a long-term perspective ( -2016 33

Source: own work
The emergence of other tourism centres of national and international interest, capable of balancing the current spatial distribution of the tourism supply, is necessary in order to discuss about a truly dispersed pattern of tourism. These findings confirm the results of the study by Constantin & Reveiu (2018), which emphasises the uneven territorial distribution of tourism infrastructure in Romania.

Different types of territories, different trajectories of tourism supply and demand
This section inquires the particularities of tourism evolution with reference to various types of territories. It does so by analysing tourism evolution for three different types of territories (leading, intermediary, and lagging) at three different geographical levels (NUTS2, NUTS3, and LAU2), in order to find out if tourism could play a role in reducing development inequalities between leading, intermediary, and lagging regions. Such an analysis could highlight areas where interventions are needed and provide useful insights in defining appropriate measures for each particular type of territory. At the regional level, the most economically developed NUTS2 regions (leading areas) also represent the territories with the most significant share in national tourism. These areas stand out through the highest values of both tourism supply and demand (Figure 3). However, regarding the accommodation establishments, the intermediary and lagging regions manifest the highest growth rates for the 1990-2016 period, which indicates a growing interest in tourism development in these cases. In the case of the values of overnight stays, all the regions register constant values over time, except for Bucharest-Ilfov region, that exhibits a noticeable growth tendency compared to the year 2001.
At the same time, at the county level, the leading areas are noticeably better represented from a tourism point of view (Figure 3). However, the lagging counties manifest a strong tendency towards increasing the accommodation capacity. Thus, for the period 2000-2016, lagging regions mark significant higher rates of growth in the number of accommodation establishments than intermediary and leading ones.

Source: own work
At the local level, a different reality concerning tourism's trends is revealed. The most significant rates of growth for all tourism indices at the local level are specific to the peripheral LAU2s (Figure 4). The peripheral areas also stand out in the case of the number of accommodation establishments. Although they are still not as developed from this point of view as the central urban areas, the profound rural administrative units register significant high numbers for the accommodation establishments. In this context, one could reasonably ask if EU's incentives and national programs aimed at encouraging tourism development in rural area explain such trends.

Source: own work
The spatial analysis of tourism supply and demand evolution by type of territory highlights lasting disparities between the lagging and leading territories. Leading regions and counties take the lion's share of the Romanian tourism growth, especially when it comes for tourism demand. Thus, tourism is not (yet) contributing to a more balanced development of Romanian regions (NUTS2) and counties (NUTS3). However, at the local level (LAU2), tourism's potential to contribute to a balanced development of the territories appears to be significant.  Two distinct periods were taken into consideration, in order to analyse and discuss the potential of the accession into the EU in influencing tourism trends. Both periods analysed (2001-2007 and 2007-2016) manifest considerable rates of growth for tourism accommodation establishments. Nevertheless, in the period 2007-2016, the geographical dispersion of tourism accommodation infrastructure was consistently more accelerated compared to the period 2000-2007.

Territorial dynamics of tourism supply before and after the EU accession
The percentage of LAU2 entering the tourism accommodation market for the first time is even higher than in the previous period: 343 LAUs are entering the tourism market between 2007 and 2016, compared to 209 in the period 2000-2007. This tendency brings substantial benefits to the Carpathian area, and in 2016 most of the LAU2s in this area register tourism activity (Figure 6). At a national level, from 2007 to 2016, the positive evolution of tourism supply characterizes 68% of the administrative units with tourism activity in the country. It is, therefore, essential to take into consideration the fact that when Romania became a member of the European Union, in 2007, a positive trend already characterized the tourism sector.
The significant tendency of growth of tourism supply after 2000 is mainly a reflection of the accelerated development of rural tourism ( Figure 5). Therefore, the emergence of tourism in many of the administrative units with no tourism activity at the beginning of the period consists in most cases in a reduced number of small-sized accommodation establishments, specific to rural tourism. The growth in the number of accommodation establishments is considerable, whereas the level of development of tourism supply is still low in many cases.

Source: own work
There is an average of 30-beds capacity in 2007 and 46-beds capacity in 2016 for all the LAU2 that entered the tourism market in each of the two periods of time. Geographically, the growth occurs either isolated, in areas with no previous tourism accommodation facilities, or in the proximity of areas of high tourism attractiveness ( Figure 6). The development of rural tourism as one of the central defining features of the growth of the tourism industry made the Romanian tourism context considerably more favourable for implementing EU policies regarding tourism development.
In the same period, a significantly lower number of LAU2s was characterized by decline tendencies. Rural and/or mountain areas were the most affected territories, but presumably, those particular administrative units presented little interest for tourism, being characterized by a poor tourism infrastructure, both in quality and quantity. During the period 2001-2007, tourism supply was declining in the case of 21% of LAU2s, while in another 13%, the tourism supply completely disappeared by the end of the period 2000-2007 ( Figure 5). Even less LAU2s faced decline between 2007 and 2016, around 8% of LAU2s being confronted with a complete disappearance of accommodation establishments. Another category of declining territories in both periods of time are the urban areas with little or no tourism value, that were the subject of tourism development during communism, based only on administrative and political reasons. Therefore, the disappearance of the accommodation establishments in these small-sized cities was a natural consequence of the prolonged and difficult process of privatization. The Seaside and Danube Delta area is another type of territory that was profoundly affected by a declining slope, between 2007-2016, with approximately half of the local administrative units facing a decrease in the number of accommodation establishments ( Figure 5).

Territorial dynamics of tourism demand before and after the EU accession
Tourism demand increased significantly after 2000. This trend was in accordance with the general trend of economic growth and, also, of tourism growth at the level of the infrastructure. 68% of the administrative units with tourism activity in the period 2000-2007, registered a positive trend. Out of these growing territories, almost 31% were cases of LAU2s entering the tourism market for the first time (Figure 7). In the period 2007-2016, the growth continued, with the European Union accession bringing significant changes in terms of tourism demand. In 2008, the number of foreign tourists increased by 15% compared to 2007. Their countries of origin covered almost the entire European continent. However, almost 60% of those came from neighbouring countries (Ilieș et al., 2017). Contrary to the general decline of tourism demand in 2009, in the context of the economic crisis, the number of foreign tourists continued to increase, in a more reduced rhythm. In this case, the crisis itself can be perceived as a cause of this growth, as Romania might have appeared in this context of economic instability as a financially more accessible destination at the European level.
Nevertheless, the positive economic context in Romania also determined side effects with a less positive impact on tourism demand. In this period of time, a preference for abroad destinations emerged among the population in Romania, with the most frequent options being the neighbouring countries (Rădulescu & Stănculescu, 2012). The factors for this behaviour are growing accessibility to foreign countries for Romanians, from a financial point of view, but also a quest for better quality in tourism services and infrastructure, as an alternative to the inferior quality that characterized most of Romanian destinations. The incidence of LAU2s that registered a decline in the tourism demand is less significant for the period 2007-2016 than for 2001-2007 (Figure 8), with a difference of almost 6% between the two periods. In the period 2007-2016, most of the tourism areas in decline correspond to destinations deeply affected by the 2008 economic crisis. Despite signs indicating positive trends in the last years, the impact of the crisis is still evident, as tourism demand is still considerably lower in 2016 compared to the time prior to the crisis. The northern part of Apuseni Mountains is such an example, with Sâncraiu, Băișoara and Beliș as cases of destinations that still have not reached, until 2016, the level of tourism demand registered before the economic crisis.

Conclusions
The analysis of the territorial dynamics of tourism in Romania revealed a series of characteristics that might prove useful from a policy perspective. The main aspect of considerable relevance is the fact that, despite a stated interest in encouraging tourism's spatial diffusion in Romania and despite evident progress in this direction, this diffusion presently has a limited observable impact. More precisely, this diffusion translates into an increase in the number of administrative units that register some form of tourism activity, but in most cases, the dimensions of this activity are of reduced importance. This situation, alongside a constant and significant growth of tourism in already solidly developed tourism destinations, determines a persistence of a concentrated pattern of tourism in Romania, with a small number of territories that lead the tourism industry in the country.
The image of tourism's territorial dynamics at different geographical scales reinforces the idea that one cannot regard tourism as a uniform activity across the Romanian territory. The imbalanced spatial distribution of tourism is strongly influenced by the economic and demographic nature of the territories. The economic leading counties and regions and the high-population-density LAU2s are the ones that stand out as main destinations in the Romanian tourism industry. They are also the ones that take the lion's share of the growth in tourism demand, transforming tourism into a factor that is more likely to increase territorial disparities than to contribute to their reduction. However, the potential of the lagging territories for tourism development becomes visible through the most significant growth rates for the supply indicators, especially in the case of peripheral LAU2. It might be a matter of time and of policies directed towards this issue for tourism demand to manifest the same degree of growth in these regions. At the same time, diminishing the difference between leading territories, and lagging ones might be an objective to be considered more in such policies, as the potential of the last category for tourism growth is evident. Therefore, it can be concluded that particular attention could be given to differentiate public policies and investments in tourism according to the type of territory.
The accession to the European Union does not appear as generating significant shifts in the trends regarding tourism territorial dynamics in Romania. However, the favourable economic and political context it created might be regarded as one of the factors that sustained and stimulated the already existent growth tendencies and of tourism spatial diffusion that started in 2000. An aspect particularly associated with the EU integration might be the interest and engagement for developing rural tourism, reflected both in the values of tourism supply and in those of tourism demand in rural areas.
This study emphasised the current tendencies of tourism development in relation to the type of territory. Further studies could search for factors explaining these tendencies or inquire about the relationship between the way tourism evolves and the level of economic development of each type of territory.