**1. Introduction**

The inclusion of the space variable in economic analyses is becoming an increasing habitual practice, especially in the case of activities, which, owing to their very nature, have a close relationship with their development in a given territory, as with tourism. As pointed out by Sánchez [1], a tourist destination cannot be analysed in isolation without taking into account the influence on it of proximal destinations and vice versa. The presence of a tourist business in a certain location will determine aspects as essential as the resources available; these are understood to be the presence of attractions and their exploitation [2,3], as at the same time this affects the occupation level, that of seasonality, and that of competitive intensity, among others [4]. Moreover, the satisfactory progress of tourist activities will also be influenced by factors such as accessibility or the supply of accommodation or complementary services [5].

The ways in which the location of an establishment in a given geographical area may influence the satisfactory progress of the activity are therefore diverse. For this reason it is not surprising that the conceptualisation of distance must be included in the statistical analyses to be performed in order to obtain an exhaustive vision of the tourist situation, and this is possible thanks to the application of the techniques of spatial statistics. This statistical tool is characterised by going further than conventional statistical analyses, including the space variable and spatial relations gathered by means of the design of a matrix of spatial weights as another parameter to be taken into account in the analysis to be performed; its use in social science is becoming more and more frequent.

This growth in the use of techniques of spatial statistics has not occurred in isolation but has been encouraged by the greater dissemination of geographic information systems (GIS) in the field of economic analysis in general and in particular, in the case of tourism. As pointed out by Anselin [6], the application of techniques of spatial statistics together with GIS extends the limits of the types of analysis that may be carried out in a realistic environment, such as those orientated towards supporting the analysis of policies or the making of decisions. It is therefore possible to synthesise the factors which have had an influence on the fact that the space variable is becoming more important in social science in the following ways: the greater importance of spatial interaction in social science, the greater availability of georeferenced databases, and the development of GIS software including specific modules allowing the statistical analysis of spatial data [7].

For all these reasons, GIS are beginning to be recognised as valuable tools for arranging, analysing, and expounding large volumes of data for any local and regional planning activities; their use is becoming imperative in tourist planning and managemen<sup>t</sup> [8]. The main advantage of the application of GIS technology in tourist analysis is that it allows the acquiring of greater knowledge of the structure and the operation of the tourist system in a given area, either for the purpose of planning or with the objective of monitoring the development of the existing activity [9].

The analysis of the distribution patterns of the variables related to tourism, identifying whether the variables tend to be concentrated or dispersed in the space, the finding of groups with characteristics similar to those of proximal locations, or on the contrary, the finding of observations of behaviour clearly di fferentiated from that of their neighbours, are some of the possibilities of spatial analysis. In other words, by means of spatial statistical analyses, two important spatial e ffects can be observed: dependence or autocorrelation and spatial heterogeneity, which will have important implications for tourist management.

The main implications of the finding of interdependence relationships in space lie in the geographic spillovers associated with them. This interdependence between regions has been analysed under the agglomeration economies approach [10–12], which is based on the premise that the concentration or spatial proximity of economic activities can be beneficial due to externalities of the agglomeration for the whole economy as well as for the sectors and companies grouped in a particular location, highlighting the improvement of productivity, investments, labor market, knowledge transfer, among other aspects [13]. In this line, the works carried out by Majewska [14,15] and Majewska & Trusklolaski [16], which, based on this premise, analyse the geograhics spillovers of tourism activity in Poland and countries of central Europe, identifying the existence of di fferent hot spots that represent essential knowledge for proper planning of these destinations.

In the same way, the works carried out by Yang and Wong [17] that, with the object of study being China, identified the presence of di fferent hot spots in coastal areas, mountainous regions, gateway cities or higher-hierarchy cities within the country that extend its e ffects beyond natural borders, as well as the existence of certain areas that constitute cold spots with a low level of tourism development. Alongside these, other works carried out for the same purpose stand out: the identification and description of the spatial pattern of tourism activity in particular territories [1,5,8–32].

As a general conclusion of all these works, it can be confirmed that the distribution of tourist activities in a region is not homogeneous [5,9,20,25,27,29,33,34]. On the contrary, this spatial distribution is characterised by a series of patterns which must be identified and taken into account for the correct managemen<sup>t</sup> and planning of a destination.

In general terms, in analysing the distribution of the demand from travellers in the territory, it can be expected that there will be a certain preference for those locations which have the most tourist attractions. In accordance with this premise, it can be anticipated that tourist lodgings have a greater supply of beds in those locations which are more attractive to the demand, or what amounts to the same, that the beds are concentrated in the locations of greater preference of the demand. This has been the approach used by a large proportion of the studies carried out to date to analyse the distribution of tourist activity in space, i.e., assimilating that supply and demand have the same distribution pattern. Therefore, the lack of specific details allowing the analysis of the behaviour of travellers may be compensated for by a detailed study of the behaviour of the supply [1,5,9,25].

It must, however, be taken into account that the creation of beds for tourists is not always a response to the prior existence of interest from the demand. This could be the case of the creation of beds in locations which lack a strong tourist tradition but which see in the development of this activity a good opportunity to generate wealth and employment.

This paper aims to investigate the degree of adjustment between supply and demand in a territory where, due to the particularities of the growth model implemented based on expansive policies, it has given rise to a strong imbalance between supply and demand, the region of Extremadura [1,5,25,28,32] that needs to be studied and analyzed in order to implement the appropriate strategies to achieve growth sustainable tourism in the region.

The principles on which sustainable tourism development rests based on various institutional declarations of the World Tourism Organization (UNWTO) according to Cánoves, et. al [35] can be synthesized as follows: giving optimum use to environmental resources, maintaining and helping to conserve natural resources and biological diversity, respect the cultural authenticity of host communities, conserve their cultural and architectural assets and their traditional values, and ensure long-term viable economic activities that benefit all agents and report socio-economic benefits. Therefore, the mere creation and provision of housing capacity is not enough, but for a truly sustainable economic system to be constituted, it is necessary that the distribution of these be adjusted correctly to the preferences of the demand, or what is the same, that these places obtain a su fficient occupancy rate to generate an economic benefit that allows the continuity of the business without exceeding maximum saturation levels that hinder the correct conservation of the main tourist attractions.

In this sense, the analysis of the occupation level seems to be a good option to generate knowledge from which to design development strategies that allow fulfilling the necessary objectives so that the sustainable tourism development of the area under study is possible.

Moreover, as some of the previous studies underline, one of the main weaknesses of the technique used is that the use of administrative boundary a ffects the results obtained [14,15], so some of the geographic spillovers could be covered up; in order to avoid this problem, it has been decided to use a global positioning system (GPS) coordinate reference unit for each of the observation units, that is, this analysis is performed from a territorial perspective disaggregated at the highest possible level, the very location of each establishment. This study uses in total as a sample, a set consisting of 270 accommodation establishments for which data are available on their beds and occupation levels for July 2015; all these establishments are located in the region of Extremadura. The use of the establishment itself as a unit of analysis will allow the identification of a set of lodgings with similar behaviour (spatial clusters), which in view of the managemen<sup>t</sup> of the territory, allows the definition of joint planning strategies.

The novelty of the approach used in this article must be sought in the combination of the methodology and destination used. Most of the work done to date to analyse the distribution pattern is carried out at an intra-urban scale [36,37], also selecting destinations that are in the maturity phase. The peculiarity of the analysed destination lies in the fact that it is an emerging interior destination that, due to the characteristics of tourism products with development potential in these destination, requires sustainable tourism management. Therefore, using an analysis focused on the e fficiency of the territory measured through the occupation level as a proxy indicator of tourism pressure is an important temporary spatial tool to locate possible locations that could present problems of excess load capacity. In the same way, it is essential to establish the appropriate development policies for the identification of possible locations that in the space and time analysed make less e fficient use of their available resources. Therefore, it is considered that the analysis performed is a valuable tool for public and private managers in order to manage the destination that allows its sustainable development.

In order to achieve its objective, this study is structured as follows: after this introduction, the next section details what is meant by exploratory spatial data analysis (ESDA) and what has been its

application in the field of tourism. The third section serves as a guide to the reader in the enumeration of some characteristics of the geographical scope of this study, the region of Extremadura. Subsequently, we describe in detail the methodology used in this research. Section five describes the results obtained, and finally this research is completed with a synthesis of the main conclusions and implications for the managemen<sup>t</sup> of the results.

#### **2. The ESDA and Its Application to the Tourist Sector**

The tourist industry is characterised by a growing need for planning, which in its turn requires techniques capable of monitoring and analysing the flows of tourists [38]. One of the characteristics traditionally attributed to tourism is its territorial dimension, which is also characterised by an unequal distribution within and between destinations [27]. For this reason, finding out the distribution pattern of the data in a given territory is an essential task in the field of tourist managemen<sup>t</sup> and planning.

The ESDA is a good tool for this purpose when no clear signs are present in the distribution patterns of a variable. The ESDA can be defined as a set of techniques which describe and visualise spatial distributions and at the same time identifies atypical locations (spatial outliers), discovers schemes of spatial association, groupings (clusters), or hotspots and also suggests spatial structures and other forms of spatial heterogeneity [39].

The importance of GIS in this process lies in the possibility which GIS o ffer for statistical analysis using graphs, which gives rise to what some authors term "scientific visualisation" [40].

By means of the exploratory study of spatial data, the so-called spatial e ffects can be identified: autocorrelation or dependence and spatial heterogeneity. Spatial heterogeneity can be defined as the variation of relationships in space. It is determined either by the presence of structural instability caused by the lack of stability in the space of the behaviour of the variable under study or by the presence of heteroscedasticity [41]. The second of these e ffects, the so-called spatial dependence or autocorrelation, is confirmed when there is a relation between what occurs at a given point in the space and what occurs in other points of the same space [42]. It is therefore in line with the contents formulated in the "first law of geography" of Tobler [43], according to which "everything is related to everything else, but near things are more related than distant things".

This study of dependence or spatial autocorrelation may result in three possible scenarios.

The first of these consists of the finding of the lack of spatial autocorrelation, i.e., when it is confirmed that the values of the variable are distributed at random in the territory analysed (random pattern of distribution).

The second of these scenarios is associated with the confirmation of positive spatial autocorrelation. This occurs when there is a direct relationship between similar values of a variable. It implies that the presence of a given phenomenon in a region means that it extends to other nearby regions [44]. In the specific case of tourism, the presence of this type of autocorrelation involves the presence of similar values of the tourist variable among nearby destinations, which means that a "contagion" e ffect therefore exists [1].

Finally, the third possible scenario is when the presence of negative spatial autocorrelation is confirmed. This occurs when nearby destinations present very di fferent values of a variable, or what amounts to the same thing, when the presence of a phenomenon in a region prevents or hinders its appearance in neighbouring regions [44]. In the specific field of tourism, this case generates what is known as the e ffect of the "absorption" of a given geographical space [1].

Moreover, it must be taken into account that the study of spatial autocorrelation can be approached from two di fferent perspectives: at a global level or a local level. The contrast from a global perspective pursues the objective of identifying spatial trends or structures in a specific geographical space, including the total of the observations of the variables in said space. In order to do so, the indicators proposed by Moran [45] and Getis and Ord [46] will be used. The main di fferentiating characteristic of these contrasts from local contrasts is that they allow the summarising of a general scheme of dependence on a single indicator [41].

For its part, the local contrast of spatial dependence or autocorrelation is characterised because an indicator is calculated for each of the observation units, owing to which they allow the identification of in which of them higher (or lower) values than those expected in a homogenous distribution are concentrated. The most popular indexes for confirming the presence of local spatial autocorrelation include the local indicators of spatial association (LISA) proposed by Anselin [47] and the Gi family of statistics of Getis and Ord [46] and Ord and Getis [48].

Although both tests can be considered complementary, their approach can be clearly di fferentiated. While the Getis–Ord Gi test concentrates on locating groupings of similar high or low values of the variable which are in accordance with the values of their neighbouring locations, Anselin's local I test expands these results to locate not only these two types of groupings but also those other entities presenting anomalous values compared with those taken by their neighbouring locations. This test may therefore give rise to five di fferent results: groupings of high or low values with neighbouring locations taking similar values (HH or LL), high-value groupings surrounded by low values (HL), low-value groupings surrounded by neighbours with high values (LH), and finally entities in which no significant relationship can be identified.

As can therefore be seen, the results obtained by the application of Moran's local I test enriches the analysis and it is for this reason that this option has been selected to perform the analysis of local spatial distribution in this study.

On the other hand, the joint use of both types of contrasts, local and global, will allow the obtaining of exhaustive results in the spatial analysis carried out. In this sense, several authors point out that these are complementary techniques, as one of the main limitations of global autocorrelation tests is their incapacity to detect local spatial structures, hotspots or coldspots that may or may not extend to the global pattern structure [44,46,47,49–52].

At the same time, both types of test, local and global, have a common problem which must be correctly approached and resolved prior to the application of these techniques; deciding what will be the conceptualisation of the relationship of proximity, i.e., how to distinguish which entities are to be considered neighbouring. In order to do so, various criteria have been established, which in turn will vary depending on the approach used: the lattice perspective or the geostatistical perspective.

In the specific case of the geostatistical perspective, which will be the approach used in this research, the relationship of proximity can be established by means of any of the following criteria: inverse distance, square inverse distance, and fixed distance band. In general terms, the fixed distance band criterion is the most frequently used in the existing literature [53,54]. Authors such as Sánchez et al. [28] point out that it should be taken into account that each of the criteria listed establishes a relationship of proximity which has a significant e ffect on the results obtained; owing to this it is necessary to be cautious and carry out di fferent tests before deciding on a criterion so as to ascertain which is best suited to the study area.

With regard to the possibilities of the application of these techniques to the specific field of tourism, it should be pointed out that this will allow the identification of the distribution pattern followed by the variable in the area analysed, with the implications for tourist planning which these findings involve. In this way the identification of a positive spatial autocorrelation pattern in a given region indicates the existence of a contagious e ffect among neighbouring destinations, which would make possible the existence of a common strategy for tourist development in neighbouring regions. At the same time, the existence of spatial autocorrelation at a local level will allow the identification of groups of municipalities with common characteristics regarding their tourist situation and therefore with similar needs as to the designing of strategies for future development.

The following section provides a series of characteristic features of the evolution of the tourist sector in Extremadura together with a reference to some studies carried out in this region which allow the reader to obtain further knowledge on the tourist situation of the region.

#### **3. Case Study: Space and Tourism in Extremadura, Spain**

Extremadura is a Spanish region in the southwest of the Iberian Peninsula which consists of the two largest provinces in Spain: Cáceres and Badajoz. The total surface area of the region is 41,633 km2. Its economy has traditionally been characterised by a strong dependence on agro-forestry activities and by being that of the Spanish region with the lowest gross domestic product (GDP) per capita [55]. Moreover, this is combined with a high rate of unemployment which was 19.68% in the third quarter of 2019 [56]. Given this situation, the region has been obliged to create new productive activities to provide economic development by means of the creation of wealth and employment and has seen in the development of tourist activities a good ally to achieve this.

The potential of tourist activities for contributing to economic development, reducing regional asymmetries, creating employment, and generating positive external elements affecting other economic activities has been traditionally accepted [57]. This characteristic becomes particularly relevant in the case of regions which owing to a geographically isolated location see in the tourist sector their only possibility of growth [58] by means of the diversification of the existing incomes.

It is as a result of all this that at a European level a series of programmes have been developed with the ultimate aim of the diversification of economic activities in areas with a low level of economic development. The LEADER, LEADER II, and PRODER programmes are a good example of this, and their impact is particularly noteworthy in the region under study, Extremadura.

The result of these subsidies has been the rapid growth of the accommodation capacity of the region, especially in the case of rural lodgings. This growth has not been matched by a parallel increase in the number of visitors, which has therefore created considerable imbalances which must be analysed exhaustively in order to understand the current situation of the tourist sector in the region [5].

In the year under study, the region had a total of 1296 accommodations that offered a total of 38,940 places. Of these, a total of 19,837 places were offered by the 471 hotel accommodations installed in the territory, the rest, 19,103 places, are the result of the offer of beds made by the 827 non-hotel accommodations located in the region in that year according to data provided by the Tourism Observatory of Extremadura. We find, therefore, a region that keeps a good balance between the number of places offered by hotel and non-hotel accommodation. However, the own peculiarities of extra-hotel accommodations, which offer on average a smaller number of places, entail a greater representativeness with respect to the number of accommodations, which in turn allows a better distribution throughout the territory.

The studies which have been carried out to date in order to find out the pattern of distribution of tourist activity in this space have taken as a reference variable the number of beds available, with these studies being limited to the analysis of the beds offered by rural tourism lodgings. In this way it has been possible to confirm the existence of different clusters of municipalities offering a high number of beds in comparison with what would be expected in a homogeneous distribution of the variable [1,5,25,28].

There is no doubt that these studies have helped to stress the importance of the analysis of spatial distribution patterns of tourism in the region as their results confirm the existence of groups of municipalities with a similar accommodation capacity and which can therefore develop common strategies of development and planning. As is stressed by the authors themselves, the existence of these similarities among territories allows a bid for joint policies to cover extensive proximal territories [5].

However, up to now, the findings on the spatial behaviour of the tourist supply in the region have only served to emphasise the need for carrying out exhaustive spatial analyses of tourist activities and in particular to provide information on the behaviour of the demand from travellers to the region and its adjustment with respect to the places offered.

One of the common conclusions of the studies carried out to date on the distribution of the beds for tourists in the region is that the expansive policies used as strategies for developing the sector have given rise to unequal growth between the supply and the demand, with the result being the generation of imbalances in the activity [1,5,25,28]. This situation of imbalance appears to be preventing these beds from fulfilling their function of generating the economic growth for which they were created.

This imbalance may be represented by one of the variables habitually used in order to characterise the satisfactory performance of tourist activities, the occupation level. The occupation level of an accommodation establishment can be defined as the quotient between the beds which have actually been occupied in relation to those available. This variable can therefore be considered a good proxy indicator of the level of adjustment between the supply of and the demand for tourist activities in a given territory. The spatial analysis of the occupation level variable will allow us on the one hand to find out whether there is a general pattern of the grouping of the variable in the space and on the other to discover groups of lodgings showing similar behaviour in the space, i.e., having a satisfactory adjustment in their supply of beds (hotspot) or on the contrary, poor adjustment between supply and demand (coldspot).

The importance of the findings which we aim to discover with the performing of this analysis lies in the fact that as a consequence of the same it will be possible to group together those accommodation establishments which are near to each other and show similar behaviour; this in turn will allow the generalisation of the tendency identified in the territory in which they are located. In short, the groupings identified by this kind of analysis will permit the regional administration to establish common development strategies, making use of the synergy e ffect and the consequent scale economies, and designing joint planning to cover extensive territories with a similar characterisation and initial situation.

Given that the objective of this research is to help the regional administration to identify territories in which tourist lodgings are located that have an equal occupation level di fferent from that to be expected in a homogeneous distribution of the activity, we decided to use as a reference the territorialisation created by the regional administration for the strategic planning of the region. Since the establishment of the Extremadura Observatory Tourism in 2013, the regional administration responsible for tourism has opted to use a territorial division that allows combining tourist regions that, because they have similarities in terms of their portfolio of tourism products, have been considered optimal to perform a joint analysis. It should be noted that this division is carried out on the basis of knowledge of each of the regions subject to territorialization but that it is not based on any study that has used ESDA techniques that have allowed us to verify the spatial grouping of accommodation whose behavior is similar and di fferent from that expected under a homogeneous distribution pattern, and that therefore, supports the feasibility of using joint planning that allows optimizing the results of the policies implemented for joint development.

In the same way, the sample used to perform this analysis is that proposed by the Extremadura Tourism Observatory and guarantees the correct representation of the supply of beds existing in each of the territories used. Specifically, for the reference month and year, the sample consisted of a total of 270 lodgings that o ffered 15,966 places in the region. According to the total accommodation capacity indicated above formed by 38,940 places distributed among a total of 1298 accommodations, the sample obtained represents 20.8% of the establishments and a total of 41.0% of the total places. In addition, it should be noted that, as specified in the di fferent reports published by the Tourism Observatory, the representativeness of the sample was determined for each of the territories, and given that sample is the sample used by the agency that is responsible for carrying out the o fficial statistics of the region, and it is confirmed, the selected sample is considered representative of the tourist activity of Extremadura.

Figure 1 shows both the distribution of the sample of accommodation establishments used and the location of each of the territories which will subsequently be analysed.

Once the sample to be analysed has been presented, the following section explains in detail the methodology used to achieve the proposed objectives of this research.

**Figure 1.** Distribution of the sample in the tourist territories. Source: Own material from calculations made with ArcGIS ver. 10.3.
