The State-of-the-Art of Smart Cities in the European Union

: Today, policymakers struggle to obtain information from speciﬁc smart city case studies. The literature lacks a uniﬁed view of current initiatives. This paper performs an empirical study with the aim of collecting evidence from the literature about existing smart city initiatives in the European Union (EU). The contribution of each paper and its geography are analyzed using content analysis to identify the number and type of initiatives in each country. A cluster analysis is performed to ﬁnd relationships between countries and their development phase as well as the categories (areas) they are focused on. The results suggest that there are different levels of smart city development between the member states despite the initial year of their ﬁrst result in the literature. Furthermore, 22 smart city categories clustered in four different groups were found. When compared to countries’ socio-economic characteristics, the results suggest the development of smart cities is signiﬁcantly related to the public budget balance, gross domestic product and EU structural funds. In summary, this paper portrays the state-of-the-art of smart city initiatives in the member states of the European Union. Moreover, it represents a valuable contribution to decision makers to discuss ways to standardize smart city approaches in the European scope. Furthermore, the method used in this paper can inspire the development of collaborative dashboards for the exchange of best practices and data accessibility about case studies’ details.


Introduction
Smart cities emerged in the late 1990s, motivated by the vision of decision makers who have the capacity to simulate the urban environment in real time [1][2][3].This vision was initially biased by the interests of technological companies to attach cities' strategies to their interests [4][5][6].It has led cities to test solutions and implement pilots without a strategic direction.This challenged the existence of case studies in the literature and the exchange of best practices between cities.Although the concept has evolved to consider cities' sustainability, the quality of life of inhabitants and their involvement in the co-creation of strategies [7][8][9][10], policymakers have been struggling to access information on specific smart city implementations.This information exists but is spread across the literature.Thus, the literature lacks a unified view of the state-of-the-art in terms of current initiatives.
The creation of a successful smart city strategy depends on existing data and correct benchmarking [11,12].Since the increase in attention to the topic in 2010 [13], there has been an exponential growth of publications with a strong multidisciplinary nature in their subjects, dominated mainly by China, Italy, USA, Spain, and England [14].However, no attention has been given to smaller countries which tend to have lower bibliometric indicators, less publications and a lower number of citations in the literature [15].Correia, Teixeira and Marques [16] reviewed the state-of-the-art of smart cities in Portugal through content analysis of a Portuguese smart cities magazine (the leading journalistic source dedicated to the topic, distributed to all Portuguese cities) and found that the literature had only a small subset of existing initiatives.The authors concluded that cities with an integrative strategy usually have national or European funds, which explains the fact that cities are very dependent on public calls to support smart city approaches.In addition, it also explains the existing heterogeneity between cities of the same country and of different countries.Furthermore, through this magazine, Portuguese policymakers had only information about few flagship cities and case studies, namely Amsterdam, Barcelona, Bristol, Budapest, Columbus, Copenhagen, Coruña, Curitiba, New Deli, Dublin, Hamburg, London, Madrid, Medellin, Moscow, New York, Ontario, Riyadh, San Francisco, Santander, California, and Songdo.Thus, even from non-scientific sources, the available information for decision makers is scarce.
In this line of thinking, Ruohomaa et al. [17] point out that research on smart cities usually focuses on the transformation of big cities, with the topic being widely neglected for medium and small cities.Several examples can be found in the literature.Ojo, Curry and Janowski [18] present the findings and lessons from a study of ten smart city programs (Amsterdam, Malmö, Malta, Masdar, PlanIT Valley, Singapore, Curitiba, Songdo, Tianjin, and Yokohama).Chang and Kuri [19] analyze four case studies (Barcelona, Curitiba, Singapore, Hammarby Sjöstad) to merge features that are commonly fragmented, and created a broader perspective for designing strategies for urban renovations and sustainable development in Latin American cities.Angelidou [20] identifies 10 distinctive characteristics of smart cities and analyzes their presence in the smart city plans of 15 cities based on studies published in the literature (Amsterdam, Barcelona, London, PlanIT Valley, Stockholm, Cyberjaya, Singapore, King Abdullah Economic City, Masdar, Skolkovo, Songdo, Chicago, New York, Rio de Janeiro, and Konza).Giest [21] studied five cities (Copenhagen, London, Malmö, Oxford, and Vienna) to conclude that the complexity of big data integration limits the capacity of local government to set up data management structures, which is why current solutions focus on local pilot sites and outsourcing data analytics to private companies.Based on evidence of European projects, Camboim et al. [22] identified the elements and features to develop a smart city strategy and stressed that the challenge lies in defining how to coordinate these driving elements in each dimension to develop the urban innovation ecosystem (Amsterdam, Barcelona, Vienna, and Lisbon).Noori, Hoppe and De Jong [23] classified the pathways for smart city development according to design, governance, and implementation (Amsterdam, Barcelona, Dubai, and Masdar).Csukás and Szabó [24] positioned nine cities into four clusters (Environmental Efficiency, Quality of Life Applications, Citizen Engagement and Social Inclusion) to ultimately allocate them to four different types of smart cities depending on their focus (Amsterdam, Barcelona, London, Helsinki, New York, Vienna, Berlin, Budapest, and Moscow).
Furthermore, the characterization of current initiatives would give policymakers and researchers the knowledge about existing case studies based on the area/category and location, which would enhance the sharing of best practices and benchmarking within the sector.Thus, the aim of this paper is to portray the state-of-the-art in the European Union by identifying existing smart city initiatives throughout its member countries in the scientific literature.
To answer the research question "What are the smart city initiatives in the European Union?" a review of the literature was performed with a search in Scopus about evidence of smart city implementations in each territory to reach conclusions about the concept development phase of each country and the typology of existing projects.

Theoretical Background
This section provides an overview about the evolution of smart city initiatives in Europe and reviews the research studies that aim to define smart city categories.This literature review supports the importance of the study and discussion of the paper because it gives readers an overview about the classification of smart city categories and their associated projects.

Smart City Initiatives
Until 2010, the number of smart city studies reported in the literature was low.Only after the emergence of the smart city projects supported by the European Commission was a proliferation of writings and academic publications on the topic noted [25].Moreover, the European Commission has been supporting and investing in smart city initiatives since its early days.In 2016 there were 34 dedicated projects in the European Union (EU) [26].
One of the first smart city references was the city of Barcelona.In 2011, the city was focused on experimentation and technological transformation via the introduction of new, innovative technologies, with a view to improving the operation and management of the city in general, promoting economic growth and strengthening the well-being of citizens [27].The top-down approach benefited technology solution providers [28].Among the initiatives implemented, there were new models of service management and relationship with citizens inspired by e-government principles as well as sustainable growth projects in the categories of smart lighting, mobility and energy, the installation of the municipal Wi-Fi network, and the creation of a living lab district "22@" [27,29].
Furthermore, in 2012, Lee and Hancock [30] mentioned the existence of 143 ongoing smart city projects.Of these, 47 were located in Europe and 30 in the US [31].
In China, according to the Chinese Smart Cities Forum, six provinces and 51 cities have included smart cities in their government plans [32].
Many projects and applications can be found in the literature about flagship cities such as Santander, Manchester, and London [33,34].The references are primarily to the applications of sensors and network infrastructure for different categories.Among them are parking solutions, waste management, traffic control, air quality monitoring and Wi-Fi or IoT networks.
The capital of Finland, Helsinki, has a smart city development area, Smart Kalasatama, facilitated by Forum Virium Helsinki (FVH), which allows the implementation of agile smart city pilots with multi-stakeholder collaboration.However, it does not have a specific smart city strategy [35].
Gohari et al. [36] literature review stresses that although the Norwegian region has come as far as its European counterparts in terms of smart city applications and projects, strategies remain in the planning stages and are still very fragmented.
These two examples of developed northern European countries raise the concern of whether cities are still committing the past errors of neglecting strategic and social aspects and focusing on technological applications.
Jonek-Kowalska and Wolniak [37] studied 287 Polish cities and concluded that the implementation of smart cities in most cities has not been possible due to the unsatisfactory level of prosperity of the residents, the difficult financial condition of cities, and unfavorable demographic trends.Therefore, the priority areas are social infrastructure and human capital [38].It may also reveal the discrepancy and heterogeneity between those cities with financial support and those without, which in this case is directly associated with their size.Thus, specific city interventions are related to national and European funding opportunities, with a lack of strategic planning for becoming a smart city in the official city documents [39].Cities have embedded various forms of smart city-related projects in their digital strategies; however, only partial implementation of relevant initiatives in their daily operations is noticed [40].The associated projects are typically supported by municipalities, funded by subsidies, and implemented in partnerships which fade out after the pilot stage.Thus, scaling is widely perceived as a major concern [41].[42] studied the implementation of the smart city concept in the Czech Republic and concluded that the larger the city, the more intensive investment activity, confirming the relationship between the size and absorption capacity of European funds.Thus, cities have been focused on specific applications and themes to leverage their smart city approaches, rather than promoting holistic strategies.Examples can be found in Finland [43], Romania [44], Slovakia [45,46], Poland and Ukraine [47].In the broader picture, municipalities have been prioritizing actions, mainly regarding the development of ICT infrastructures and e-government [48].In addition, there are countries such as Hungary where the topic has still not brought meaningful change nor impacted urban policy practices [49].A contrast between the national and the local levels is also noted, even for Sweden, which chose to invest in a national digitization council to decide whether local experimentations can move forward [50].

Smart City Categories
The smart city categories have expanded over the years in terms of scope and number.Several authors have reflected on and identified the main dimensions of smart cities in the literature [51].
Mahizhnan [52] defined four dimensions: IT education, IT infrastructure, IT economy and quality of life.Giffinger [53,54] considered six dimensions: smart economy, smart people, smart governance, smart mobility, smart environment and smart living, Albino, Berardi and Dangelico [55] considered the same dimensions and linked them with aspects of urban context such as industry, education, e-democracy, logistics and infrastructures, efficiency and sustainability, security, and quality.Eger [56] referred to technology, economic development, job growth, increased quality of life.Thuzar [57] considered the quality of life, sustainable economic development, management of natural resources through participatory policies, and the convergence of economic, social, and environmental goals.Nam and Pardo [58] mentioned economic, sociopolitical issues of the city, economic-technical-social issues of the environment, interconnection, instrumentation, integration, applications and innovations.Other studies have proposed: smart health, smart security systems, smart building, smart government, smart tourism, smart grid, smart transportation, smart environment, smart home and smart lifestyle [59][60][61].Barrionuevo et al. [62] considered economic (GDP, sector strength, international transactions, foreign investment), human (talent, innovation, creativity, education), social (traditions, habits, religions, families), environmental (energy policies, waste and water management, landscape), and institutional aspects (civic engagement, administrative authority, elections).Kourtit and Nijkamp [63] stated human capital (e.g., skilled labor force), infrastructural capital (e.g., high-tech communication facilities), social capital (e.g., intense and open network linkages), and entrepreneurial capital (e.g., creative and risk-taking business activities).Chourabi et al. [64] considered management and organizations, technology, governance, policy context, people and communities, economy, built infrastructure, and natural environment.Neirotti et al. [65] presented twelve domains for urban development.Mohanty, Choppali and Kougianos [66] categorized smart cities into ten components: smart infrastructure, smart building, smart transportation, smart energy, smart healthcare, smart technology, smart governance, smart education, and smart citizens.Ahvenniemi et al. [67] considered ten sector categories: natural environment; built environment; water and waste management; transport; energy; economy; education, culture, science and innovation; well-being, health and safety; governance and citizen engagement; and ICT.Wolf et al. [68] suggested an unified framework combining some of the previous approaches, which include six different dimensions, structured on two levels (macro and micro).Level one: living (quality of life), economy (competitiveness) and environment (natural resources); and level two: institutions (governance/participation), digital systems (transport and ICT) and people (social and human capital).
After a systematic literature review, Camboim, Zawislak and Pufal [22] identified the most frequent words and categorized the main dimensions associated with the smart city concept.However, they organize this taxonomy into high-level dimensions (environ-urban, techno-economic and social-institutional) and do not match the elements with existing project categories.Furthermore, it remains unclear what the existing categories and projects by country are.

Materials and Methods
As stated before, this study focused on the analysis of the current state of the European Union.For each of the 27 countries of the European Union, evidence was sought that could portrait the scope of the existing smart city initiatives.The methodology used for the analysis of the literature is demonstrated in Figure 1.Moreover, quantitative, and qualitative analyses were performed in order to find the smart city categories and existing country clusters.
one: living (quality of life), economy (competitiveness) and environment (natural resources); and level two: institutions (governance/participation), digital systems (transport and ICT) and people (social and human capital).
After a systematic literature review, Camboim, Zawislak and Pufal [22] identified the most frequent words and categorized the main dimensions associated with the smart city concept.However, they organize this taxonomy into high-level dimensions (environ-urban, techno-economic and social-institutional) and do not match the elements with existing project categories.Furthermore, it remains unclear what the existing categories and projects by country are.

Materials and Methods
As stated before, this study focused on the analysis of the current state of the European Union.For each of the 27 countries of the European Union, evidence was sought that could portrait the scope of the existing smart city initiatives.The methodology used for the analysis of the literature is demonstrated in Figure 1.Moreover, quantitative, and qualitative analyses were performed in order to find the smart city categories and existing country clusters.Furthermore, a search on Scopus using the search keywords "Smart Cit*" + ["Name of the Country"].In total, 27 queries were performed (one for each European Union member country).From the obtained results, the titles, abstracts and keywords were analyzed to allocate a first code connected to the identification of the city, and a second code concerning the contribution of the paper.This information was updated in a dedicated database (built in a spreadsheet) where the fields "City Location" and "Category Keywords" were filled, respectively.Furthermore, in terms of eligibility (inclusion and exclusion criteria) two exclusion filters were applied.
First, repeated and non-relatable papers (that do not fit the purpose of this research) were excluded from the sample, as were non-English manuscripts.Furthermore, only journal and conference papers were considered.Furthermore, a search on Scopus using the search keywords "Smart Cit*" + ["Name of the Country"].In total, 27 queries were performed (one for each European Union member country).From the obtained results, the titles, abstracts and keywords were analyzed to allocate a first code connected to the identification of the city, and a second code concerning the contribution of the paper.This information was updated in a dedicated database (built in a spreadsheet) where the fields "City Location" and "Category Keywords" were filled, respectively.Furthermore, in terms of eligibility (inclusion and exclusion criteria) two exclusion filters were applied.
First, repeated and non-relatable papers (that do not fit the purpose of this research) were excluded from the sample, as were non-English manuscripts.Furthermore, only journal and conference papers were considered.
The second filter aimed to remove the articles that did not mention any specific city or case study.Furthermore, if there was no location in the title, abstract and keywords, the paper's contribution was considered as being on the country, since the aim of this paper was to study the current situation of the country by the existing specific projects in the cities.The same happened if it was a generic study of numerous cities in the country.Nevertheless, these results were excluded from the final detailed analysis.Furthermore, if the abstract mentioned a neighborhood or an urban district of the city, the initiative was given to that city.Double counting was only allowed when the paper referred to specific case studies in different cities, regardless of the country.In addition, each paper was only allocated to a single category.
The number of search results is illustrated in Figure 2. From an initial sample of 1665 papers, 945 results were considered in the present study (56.76%).
The second filter aimed to remove the articles that did not mention any specific ci or case study.Furthermore, if there was no location in the title, abstract and keyword the paper's contribution was considered as being on the country, since the aim of th paper was to study the current situation of the country by the existing specific projects the cities.The same happened if it was a generic study of numerous cities in the countr Nevertheless, these results were excluded from the final detailed analysis.Furthermor if the abstract mentioned a neighborhood or an urban district of the city, the initiative wa given to that city.Double counting was only allowed when the paper referred to specif case studies in different cities, regardless of the country.In addition, each paper was on allocated to a single category.
The number of search results is illustrated in Figure 2. From an initial sample of 166 papers, 945 results were considered in the present study (56.76%).

Figure 2. Number of Scopus results per European Union country after applying exclusion criteria
Of these results, on the one hand, Italy, Spain and Germany were the countries wit the most results.On the other hand, countries such as Malta, Lithuania, Luxembourg, Bu garia, Cyprus, Slovenia and Croatia showed that this subject is still in its early days.
After that, the cities were organized according to their origin countries, and an in ductive thematic analysis was performed to find the smart city categories of the initiative Finally, quantitative analysis was performed to reflect the number of results per countr and the years of publication on the one hand, and on the other hand a hierarchical clust analysis based on Ward's method was performed to find country clusters to study the relationship between the categories of the initiatives (from the qualitative analysis) an socio-economic characteristics (e.g., population and area of the territory).Figure 3 sum marizes both directions of the methodology.Of these results, on the one hand, Italy, Spain and Germany were the countries with the most results.On the other hand, countries such as Malta, Lithuania, Luxembourg, Bulgaria, Cyprus, Slovenia and Croatia showed that this subject is still in its early days.
After that, the cities were organized according to their origin countries, and an inductive thematic analysis was performed to find the smart city categories of the initiatives.Finally, quantitative analysis was performed to reflect the number of results per country and the years of publication on the one hand, and on the other hand a hierarchical cluster analysis based on Ward's method was performed to find country clusters to study their relationship between the categories of the initiatives (from the qualitative analysis) and socio-economic characteristics (e.g., population and area of the territory).Figure 3 summarizes both directions of the methodology.

Results
This section presents the results of the literature analysis.Thus, it mirrors the countries' representation according to the number of smart city initiatives.Several country

Results
This section presents the results of the literature analysis.Thus, it mirrors the countries' representation according to the number of smart city initiatives.Several country clusters emerged by looking at the existing relationships between their associated results.Furthermore, the clusters were obtained by aggregating them according to (1) the number of literature results, and (2) the category groups (from factorial analysis) of the projects involved.Ultimately these results are compared with the socio-economic aspect and discussed according to the existing literature.

Country Clusters by the Number of Smart City Initiatives
In the analysis of the relationships between countries in terms of the literature results, four different clusters emerged.The first cluster comprises the countries that have less than 25 results and is composed of Lithuania, Malta, Croatia, Hungary, Bulgaria, Luxembourg, Cyprus, Estonia, Latvia, and Slovenia.The second cluster aggregates the countries that have up to 55 results, namely Ireland, Slovakia, Romania, Finland, Greece, Austria, Belgium and Denmark.The third and fourth clusters join the countries with the highest number of results.Moreover, the third cluster is composed by the countries that have up to 130 results, namely Czech Republic, France, Sweden, the Netherlands, Poland, Portugal and Germany.The fourth and last cluster comprises Spain and Italy with more than 130 results.Furthermore, Figure 5 details the cluster analysis and its representation in the map with different colors.

Country Clusters by the Number of Smart City Initiatives
In the analysis of the relationships between countries in terms of the literature results, four different clusters emerged.The first cluster comprises the countries that have less than 25 results and is composed of Lithuania, Malta, Croatia, Hungary, Bulgaria, Luxembourg, Cyprus, Estonia, Latvia, and Slovenia.The second cluster aggregates the countries that have up to 55 results, namely Ireland, Slovakia, Romania, Finland, Greece, Austria, Belgium and Denmark.The third and fourth clusters join the countries with the highest number of results.Moreover, the third cluster is composed by the countries that have up to 130 results, namely Czech Republic, France, Sweden, the Netherlands, Poland, Portugal and Germany.The fourth and last cluster comprises Spain and Italy with more than 130 results.Furthermore, Figure 5 details the cluster analysis and its representation in the map with different colors.
Belgium and Denmark.The third and fourth clusters join the countries with the highest number of results.Moreover, the third cluster is composed by the countries that have up to 130 results, namely Czech Republic, France, Sweden, the Netherlands, Poland, Portugal and Germany.The fourth and last cluster comprises Spain and Italy with more than 130 results.Furthermore, Figure 5 details the cluster analysis and its representation in the map with different colors.

Smart City Categories
From the content analysis of the codes related to each paper's contribution (allocated in the first step), 22 smart city categories can be organized: water and irrigation;

Countries' Representation According to the Smart City Categories 4.2.1. Smart City Categories
From the content analysis of the codes related to each paper's contribution (allocated in the first step), 22 smart city categories can be organized: water and irrigation; environment and air quality; waste; energy and lighting; strategy and governance; urban planning; culture, tourism and heritage; rural and agriculture; sport; smart city foundations; digitization and interoperability; privacy, security and safety; health and well-being; infrastructure and communication networks; buildings and housing; traffic; parking; mobility and transportation; logistics; community, participation and inclusion; education; and economy and industry.Table 1 summarizes a multiple step inductive thematic analysis that produced the smart city categories (Appendix A mirrors the step before).

Smart City Initiatives by Category
The total of smart city initiatives by category is represented in Figure 6.Thus, energy and lighting; community, participation, and inclusion; and mobility and transportation are the categories with the greatest number of results.On the opposite side, the categories that have less than 20 results are education; health and well-being; waste; logistics; rural and agriculture; parking; and sport.

Representation of the Smart City Categories by Country
The initiatives sorted by category are detailed in Appendix B. Furthermore, analysis of Appendix B provides information about the cities and countries that have or had projects within each category.This provides useful information for a greater understanding of current initiatives, allowing better organization for researchers and decision makers to know the geographies where a specific problem is being addressed.Thus, the categories that are present in a higher number of countries are mobility and transportation (22 out of 27), environment and air quality (21 out of 27), and community, participation and inclusion (21 out of 27).On the other hand, the cities that have a higher representation of categories are Italy (21 out of 22) and Spain (21 out of 22), which only have no project in the category of sport, followed by Germany (20 out of 22).Furthermore, the information about the smart city categories by country is detailed below in Figure 7.This graphical representation summarizes the information of Appendix C.

Representation of the Smart City Categories by Country
The initiatives sorted by category are detailed in Appendix B. Furthermore, analysis of Appendix B provides information about the cities and countries that have or had projects within each category.This provides useful information for a greater understanding of current initiatives, allowing better organization for researchers and decision makers to know the geographies where a specific problem is being addressed.Thus, the categories that are present in a higher number of countries are mobility and transportation (22 out of 27), environment and air quality (21 out of 27), and community, participation and inclusion (21 out of 27).On the other hand, the cities that have a higher representation of categories are Italy (21 out of 22) and Spain (21 out of 22), which only have no project in the category of sport, followed by Germany (20 out of 22).Furthermore, the information about the smart city categories by country is detailed below in Figure 7.This graphical representation summarizes the information of Appendix C. Appendix C and Figure 7 enable the connection of the main initiatives to each o countries to be identified.It is evident that the categories "energy and lighting" and " munity, participation and inclusion" are those with the greatest preponderance of a initiatives developed; in the first, only three (out of 27) countries have no initiative in area/category (Ireland, Malta and Lithuania), while in the case of the second there are countries (Bulgaria, Lithuania, Malta and Luxembourg).Moreover, these are the countries with the fewest initiatives in the various categories analyzed (with the excep of Ireland, all with less than five connections).
Figure 8 uses Social Network Visualizer (https://socnetv.org/,accessed on 14 2022) to demonstrate the relationships between the categories and countries, highligh those that have a greater number of initiatives among the countries.Moreover, the gories with greater representation among the countries are energy and lighting, com nity, participation and inclusion, and digitization and interoperability.This informa can also be consulted in Appendix C. Appendix C and Figure 7 enable the connection of the main initiatives to each of the countries to be identified.It is evident that the categories "energy and lighting" and "community, participation and inclusion" are those with the greatest preponderance of all the initiatives developed; in the first, only three (out of 27) countries have no initiative in this area/category (Ireland, Malta and Lithuania), while in the case of the second there are four countries (Bulgaria, Lithuania, Malta and Luxembourg).Moreover, these are the five countries with the fewest initiatives in the various categories analyzed (with the exception of Ireland, all with less than five connections).
Figure 8 uses Social Network Visualizer (https://socnetv.org/,accessed on 14 June 2022) to demonstrate the relationships between the categories and countries, highlighting those that have a greater number of initiatives among the countries.Moreover, the categories with greater representation among the countries are energy and lighting, community, participation and inclusion, and digitization and interoperability.This information can also be consulted in Appendix C.

Country Cluster Analysis (Dimensionality Reduction)
This section presents a cluster analysis that aggregates countries according to the typology of initiatives.
Two distinct approaches were undertaken, one assuming the percentage of projects that each country carried out in each category; and the other, (ii) considering the total number of initiatives, in absolute value, in each category.However, the results of the latter approach are presented and explored in more detail since the territorial patterns are not so clear and evident when considering the relative weight of the initiatives (see the results in Appendix D).Since there is a great variety of initiatives, the importance that each type of initiative assumes in each country also varies greatly.

Country Cluster Analysis (Dimensionality Reduction)
This section presents a cluster analysis that aggregates countries according to the typology of initiatives.
Two distinct approaches were undertaken, one assuming the percentage of projects that each country carried out in each category; and the other, ii) considering the total number of initiatives, in absolute value, in each category.However, the results of the latter approach are presented and explored in more detail since the territorial patterns are not so clear and evident when considering the relative weight of the initiatives (see the results in Appendix D).Since there is a great variety of initiatives, the importance that each type of initiative assumes in each country also varies greatly.
Thus, in order to group the categories in large dimensions, a principal component analysis (PCA) was previously carried out.PCA is an unsupervised learning method that transforms the main features of a large set of variables (based on their similarities or correlation) into a smaller set of latent variables (see for more details) [69].Table 2 represents the loadings (correlation of the initial variables with the new variables) that resulted from the application of the PCA analysis and shows that the twenty-two (22) categories were grouped into four (4) large areas, with a loss of only approximately 15% of the initial information.Thus, in order to group the categories in large dimensions, a principal component analysis (PCA) was previously carried out.PCA is an unsupervised learning method that transforms the main features of a large set of variables (based on their similarities or correlation) into a smaller set of latent variables (see for more details) [69].Table 2 represents the loadings (correlation of the initial variables with the new variables) that resulted from the application of the PCA analysis and shows that the twenty-two (22) categories were grouped into four (4) large areas, with a loss of only approximately 15% of the initial information.By analyzing the high values of the loadings in Table 2, each of the factors can be characterized as follows:

•
The first factor joins initiatives related to the more tangible and material issues of the smart city, such as basic infrastructures (waste, water and air) and transport (parking, communication networks, traffic and mobility).Spain is clearly the leading country in this type of project, followed by Poland, Portugal and Italy (see Figure 9-factor 1); • Factor 2 involves sectoral initiatives linked to sport, health, education and culture; Italy stands out as the country that invests most in this type of initiative, followed by Ireland and Romania (see Figure 9-factor 2);

•
As for the third factor, a concern with the competitiveness of the entire territory is highlighted.Thus, it is possible on the one hand to identify projects related to rural and agriculture, and on the other initiatives in the category of logistics, digitalization, economy and industry.In these categories, four countries can be highlighted, firstly Germany, prominently, and then France, Italy and Poland (see Figure 9-factor 3); • Finally, and contrary to factor 1, in this factor we have four initiatives of a more intangible nature linked to community participation and issues of security and privacy, in which the Netherlands appears in a leading position, followed by Ireland and Spain (see Figure 9-factor 4).
Using the scores of the four new factors as inputs, a hierarchical cluster analysis (Ward method for aggregation criteria) was applied, allowing the construction of seven different groups of countries, as shown in Figure 10.The significance levels were calculated from the application of the non-parametric Kruskal-Wallis test.
Each cluster emerged by looking at the number of projects involved in the four major categories (factor 1, 2, 3, and 4).Furthermore, by looking at Figure 10, on the left, red and green arrows can be seen that mirror the behavior of that factor for a particular cluster.In addition, on the right there is the color-coded geographic location of each country-cluster.

•
Cluster 1 includes four countries (Hungary, France, Austria, and Poland), which are countries that have many initiatives related to agriculture, logistic, housing, digitalization (factor 3) and a few in factor 2; that is, those associated with sports, health, education and culture.

•
The eight countries (Ireland, Greece, Finland, Belgium, Bulgaria, Latvia, Cyprus, and Croatia) included in cluster 2 have tended to invest little in factors 1 and 3 (basic infrastructures, transports, logistics, digitalization, economy) when compared with infrastructure-related initiatives that have greater importance.

•
Cluster 3 (Estonia, Lithuania, Slovenia, Luxembourg, Malta, Slovakia, Sweden, Denmark, Romania, Czech Republic, and Portugal) is distinguished by low investment in projects related to community participation (factor 4) and rural and agriculture activities (factor 3).
• Clusters 4, 5, 6 and 7 are singular clusters, i.e., formed by a single member state because the values are so distinct that they appear as outliers and thus form a single cluster.In this case, the single-country clusters are Germany, the Netherlands, Spain and Italy.Those are the countries with the greatest number of initiatives in each of the major categories (factor 3, factor 2, factor 4 and factor 1, correspondingly).Using the scores of the four new factors as inputs, a hierarchical cluster analysis (Ward method for aggregation criteria) was applied, allowing the construction of seven different groups of countries, as shown in Figure 10.The significance levels were calculated from the application of the non-parametric Kruskal-Wallis test.Additionally, as stated before, instead of the typology of the initiatives, a cluster analysis was also performed considering the relative percentage of the categories in each country, represented in Appendix D (spatial patterns according to the percentage of the initiatives-dimensionality reduction and cluster analysis).However, as demonstrated in Figure A1, the results are homogeneous for a vast majority of the territory, which may not add meaningful insights to this research.The particularity of these results is that they justify the argument that there is no clear pattern of typology of initiatives when the analysis is purged of the fact of the size of the country.Even so, there is the possibility of identifying which latent dimensions underlie each of the categories.Nevertheless, they deserve a mention since they justified the decision to move forward on the analysis using the clusters by the type of initiatives, hence the option of placing these results in the appendix.Each cluster emerged by looking at the number of projects involved in the four major categories (factor 1, 2, 3, and 4).Furthermore, by looking at Figure 10, on the left, red and green arrows can be seen that mirror the behavior of that factor for a particular cluster.In addition, on the right there is the color-coded geographic location of each country-cluster.

•
Cluster 1 includes four countries (Hungary, France, Austria, and Poland), which are countries that have many initiatives related to agriculture, logistic, housing, digitalization (factor 3) and a few in factor 2; that is, those associated with sports, health, education and culture.

•
The eight countries (Ireland, Greece, Finland, Belgium, Bulgaria, Latvia, Cyprus, and Croatia) included in cluster 2 have tended to invest little in factors 1 and 3 (basic infrastructures, transports, logistics, digitalization, economy) when compared with infrastructure-related initiatives that have greater importance.

•
Cluster 3 (Estonia, Lithuania, Slovenia, Luxembourg, Malta, Slovakia, Sweden, Denmark, Romania, Czech Republic, and Portugal) is distinguished by low investment in projects related to community participation (factor 4) and rural and agriculture activities (factor 3) • Clusters 4, 5, 6 and 7 are singular clusters, i.e., formed by a single member state because the values are so distinct that they appear as outliers and thus form a single cluster.In this case, the single-country clusters are Germany, the Netherlands, Spain and Italy.Those are the countries with the greatest number of initiatives in each of the major categories (factor 3, factor 2, factor 4 and factor 1, correspondingly).
Additionally, as stated before, instead of the typology of the initiatives, a cluster analysis was also performed considering the relative percentage of the categories in each country, represented in Appendix D (spatial patterns according to the percentage of the initiatives-dimensionality reduction and cluster analysis).However, as demonstrated in Fig- ure A1, the results are homogeneous for a vast majority of the territory, which may not add meaningful insights to this research.The particularity of these results is that they justify the argument that there is no clear pattern of typology of initiatives when the analysis is purged of the fact of the size of the country.Even so, there is the possibility of identifying which latent dimensions underlie each of the categories.Nevertheless, they deserve a

Socio-Economic Relationship
From a Eurostat list of 27 indicators [70] in Appendix E, this section seeks to analyze each of the cluster typologies obtained in the previous section with some socio-economic characteristics.Thus, after applying a principal component analysis, the most relevant variables (with the highest loading value) can be selected in each component.Furthermore, they are organized according to nine carefully selected, independent indicators (see Figure 11).
These independent dimensions are organized into four major categories: (i) the first, structural revenues (A, B and C), represents the economic context of each country, evaluated by the wealth produced (total and per capita) and the public revenues generated as a percentage of the GDP; (ii) the second, conjunctural revenues (D), is reflected by the "European structural and investments funds, in this case between 2014 and 2020"; (iii) the third dimension refers to public investment in general as a percentage of the GDP (F); and finally, (iv) the fourth category corresponds to expenditure at the sectorial level, according to three levels (G: general services, H: social protection and I: education).It can be said that dimensions A, B, C and D are, to some extent, the income generated by the production of the whole country's economy and external structural supports (inputs); while dimensions F, G, H and I are each member state's expenditures or investments (outputs).In a balance between the inputs and outputs, an additional dimension is presented that reflects the public budget balance measured in billions of euros (E).
From a Eurostat list of 27 indicators [70] in Appendix E, this section seeks to analyze each of the cluster typologies obtained in the previous section with some socio-economic characteristics.Thus, after applying a principal component analysis, the most relevant variables (with the highest loading value) can be selected in each component.Furthermore, they are organized according to nine carefully selected, independent indicators (see Figure 11).These independent dimensions are organized into four major categories: (i) the first, structural revenues (A, B and C), represents the economic context of each country, evaluated by the wealth produced (total and per capita) and the public revenues generated as a percentage of the GDP; (ii) the second, conjunctural revenues (D), is reflected by the "European structural and investments funds, in this case between 2014 and 2020"; (iii) the third dimension refers to public investment in general as a percentage of the GDP (F); and finally, (iv) the fourth category corresponds to expenditure at the sectorial level, according to three levels (G: general services, H: social protection and I: education).It can be said that dimensions A, B, C and D are, to some extent, the income generated by the production of the whole country's economy and external structural supports (inputs); while dimensions F, G, H and I are each member state's expenditures or investments (outputs).In a balance between the inputs and outputs, an additional dimension is presented that reflects the public budget balance measured in billions of euros (E).
Each of the graphs in Figure 11 describe the performance of the group of countries belonging to each cluster in the nine dimensions described, and it can also be identified which ones have significant differences, in this case dimensions A, D and E (the line of the graph of these dimensions has a greater color intensity).
Previously, each of the seven clusters shown in Figure 10 aggregates countries that have the same profile of project typologies: cluster 1-Hungary, France, Austria, and Poland; cluster 2-Ireland, Greece, Finland, Belgium, Bulgaria, Latvia, Cyprus, and Croatia; cluster 3-Estonia, Lithuania, Slovenia, Luxembourg, Malta, Slovakia, Sweden, Denmark, Each of the graphs in Figure 11 describe the performance of the group of countries belonging to each cluster in the nine dimensions described, and it can also be identified which ones have significant differences, in this case dimensions A, D and E (the line of the graph of these dimensions has a greater color intensity).
Previously, each of the seven clusters shown in Figure 10 aggregates countries that have the same profile of project typologies: cluster 1-Hungary, France, Austria, and Poland; cluster 2-Ireland, Greece, Finland, Belgium, Bulgaria, Latvia, Cyprus, and Croatia; cluster 3-Estonia, Lithuania, Slovenia, Luxembourg, Malta, Slovakia, Sweden, Denmark, Romania, Czech Republic, and Portugal; cluster 4; Germany; cluster 5-the Netherlands; cluster 6-Spain; cluster 7-Italy.Therefore, as an example of interpretation of Figure 11, in terms of structural revenue, Germany (cluster 4) has a high GDP, high GDP per capita and public revenue; regarding the conjunctural scope, it has a significant contribution of EU structural funds and a low public investment when compared to the percentage of GDP; related to sectoral investments, it has a high value in general services and social protection, and medium in education.Detailed results of the socio-economic characteristics in each of the seven clusters are presented in Appendix E.

Discussion
The results show different levels of smart city development between European countries.This corroborates the Portuguese internal context analyzed by Correia, Teixeira and Marques [16].Moreover, there is a gap not just between cities and regions of the same country, but also from different countries, even if they are under the umbrella of the European Union.Although every European Union member had literature results and is therefore represented in this study, there is a significant difference between the member states.Although this discrepancy can be partially connected with the starting year of the literature results, it does not explain everything.For example, Czech Republic, Italy, Ireland, and Slovenia had their first literature result in 2013 and the total results are very different between them.Furthermore, it can be associated with the fact that there are different levels of local government engagement and prioritization of investment.Moreover, this demonstrates that although the European Union has been making efforts to develop this topic, it lacks standardization and enforcement in its approach.
Several authors had previously studied the smart city areas/categories [22,[51][52][53][54][55][56][57][58][62][63][64][65][66][67][68].However, they failed to obtain a common standard understanding because of their methodological approach.After several years of existence and evolution of the topic, the smart city categories should be defined based on an inductive thematic analysis of the initiatives present in the literature.This way it can be ensured that every category is represented.The tendency is for more categories to emerge in the future.As stated in the literature review section, the scope of smart city categories has been evolving and widening to enable a more detailed understanding.Furthermore, that can also be noted in the results obtained in the cluster analysis.In line with Giffinger [53], who considered six dimensions (smart economy, smart people, smart governance, smart mobility, smart environment, and smart living), the first component that combines parking, infrastructure and communication networks, smart city foundations, environment and air quality, waste, traffic, water and irrigation, mobility and transportation, and strategy and governance covers the fundamental and initial areas of smart cities.
In the second component can be found the categories of sport, health and well-being, education, culture, tourism and heritage, and energy and lighting, which can be seen as an extended vision of component 1.Thus, most of these categories were also named by other studies [59][60][61], in which smart health, smart security systems, smart building, smart government, smart tourism, smart grid, smart transportation, smart environment, smart home and smart lifestyle were proposed.
The categories represented in the third component (logistics, buildings and housing, digitization and interoperability, and economy and industry) are in line with the dissemination and relationship with Industry 4.0 [71,72].Furthermore, the same categories were found by Albino, Berardi and Dangelico [55] when linking the former with urban living aspects.
The fourth component (categories of privacy, security and safety, and community, participation, and inclusion) is ultimately connected with Smart City 3.0, the most recent phase of the smart city concept.Here, most of the clusters, meaning all countries, have been making efforts and creating projects under this scope.Smaller countries or those that started later have already focused on citizen engagement (skipping the first stages of the concept).
Although the results at first sight suggested that energy and lighting was the category with the most results, when looking at the number of countries per category it showed that the most representative categories were mobility and transportation (22 out of 27), environment and air quality (21 out of 27), and community, participation and inclusion (21 out of 27).This may be a good sign that countries are following the development of the concept by evolving citizens in early stages of urban planning and are aware of present-day challenges regarding mobility and climate change.Furthermore, this can be found when comparing the factors to countries' socio-economics, as in the examples of Ireland, Greece, Finland, Belgium, Bulgaria, Latvia, Cyprus, and Croatia.Nevertheless, there are countries with similar characteristics, with low GDP and a small contribution of EU structural funds such as Estonia, Lithuania, Slovenia, Luxembourg, Malta, Slovakia, Sweden, Denmark, Romania, Czech Republic, and Portugal; although they have a neutral public budget balance and high public investment, they struggle to move to later stages of the smart city concept and are focused on their infrastructure.However, countries such as Austria and Poland that also present low GDP but have a high contribution of EU structural funds and public investment are already focused on later stages of the concept.Thus, although smart city development is often associated with greater urban centers and countries with higher GDP, there are other significant factors that are crucial to cities' success.
Furthermore, Germany and the Netherlands are the only ones with a positive public budget balance (revenues higher than spendings) and a high GDP per capita, which enables them to be focused on the competitiveness of the entire territory or pass through the various stages of the concept (with significant projects in every factor).

Conclusions
This study leads to the conclusion that a positive public budget balance allows countries to focus on the entire territory.On top of this, countries with significant GDP and EU structural fund contributions were able to evolve at the pace of the smart city concept, with projects in the greatest number of categories.Conversely, countries with low GDP, small contribution from EU structural funds, a neutral public budget balance and high public investment have two different approaches.Some try to close the existing gap and move their efforts to focus on community participation, and others are still investing in infrastructure (the first phase of the concept), which may mean that the existing differences through the member countries of the European Union will continue to exist.
Furthermore, this study demonstrated the heterogeneity of smart city development among European Union member states and the lack of standardized approaches.A strategic direction is needed by the regulator to allow the definition of specific guidelines upon which countries and municipalities should base their action.This strategy should also qualify, organize, and promote the relationship between companies, academia, knowledge centers and municipalities to ensure greater adequacy and sustainability of the projects and associated research.Thus, the discrepancy between territories should be combated with specific policies promoted by the European Commission to guarantee homogeneity between countries and their cities.
This research sets the beginning of a collaborative approach towards countries' involvement in finding synergies while promoting their financial sustainability by allowing others to learn from previous mistakes.Thus, this study should be used by decision makers as reference material to benchmark existing case studies and make contact with other policymakers to exchange knowledge and best practices (considering the type and location of the initiatives).Furthermore, an open dashboard can be created to uniformize each project where relevant characteristics (e.g., problem, solution, level of investment, methodology and people involved) can be associated.In future studies, the same methodological approach can be applied to other regions and continents to study the world state-of-the-art on the topic.Countries' level of investment in innovation, the budget for each of the categories and whether there is a political background relationship with the dispersion of the number of initiatives can also be studied.
Some limitations can also be pointed out in the study.First, the keywords defined for each paper in the qualitative analysis may not exactly reflect the contribution of the paper, since the texts were not read in full.Nevertheless, the authors know that the title and abstract of a paper should state clearly the research contribution.Second, the results obtained may not perfectly portray the reality of the country, since the geography of the initiative could not be found in the abstract.Additionally, others may have not been included.However, this limitation will only serve as proof that there is a lack of documentation of case studies and practical implementations in the literature that can serve as a study of good practices.Third, the assessment of the level of smart city development in a specific country based on the number of Scopus-indexed articles may be disputable.However, since smart cities are an innovation-related hot topic, their case studies are usually found in the literature.If not, the reasons may be the lack of results and conclusion of the project and not the lack of innovative material to be published.Furthermore, research centers and universities are usually connected to these projects.

Topic Cities Countries
Environment and Air Quality Table A2.Cont.

Topic Cities Countries
Culture, Tourism and Heritage Table A2.Cont.

Community, Participation & Inclusion
Belgium (1), Czech Republic (1), Denmark (1), France (1), Germany Appendix C    1 As mentioned in the Methodology section "If there was not any location in the title, abstract and keywords, the paper's contribution was considered as of the country.
The same happened if it were a generic study of numerous cities in the country.These results were excluded from the final detailed analysis".Malta's results did not mentioned any city in particular.

Appendix D
The methodology and analysis are the same of the one described in Section 4.2.4.The difference of the Country Cluster Analysis (dimensionality reduction) performed in this appendix is that instead of the typology of the initiatives, was considered the relative percentage of the categories in each country (spatial patterns according to the percentage of the initiatives).Furthermore, the goal was to find patterns according to the weight of each category compared to the total number of initiatives (and categories) a specific country had.Thus, the categories without projects were excluded in this analysis.The aim was to find if there were similarities between countries when considering only the categories that they contained smart city initiatives.

Figure 1 .
Figure 1.Methodological approach to analyze smart city initiatives in the EU.

Figure 1 .
Figure 1.Methodological approach to analyze smart city initiatives in the EU.

Figure 2 .
Figure 2. Number of Scopus results per European Union country after applying exclusion criteria.

Figure 3 .
Figure 3. Detailed methodological approach to analyzing the literature.

Figure 3 .
Figure 3. Detailed methodological approach to analyzing the literature.

4. 1 .Figure 4 .
Figure 4 shows this is a recent topic.As stated in the literature, until 2010, the number of smart city studies was low.Only after the emergence of the smart city projects supported by the European Commission was a proliferation of writings and academic publications on the topic noted (Jucevičius, Patašien ė, & Patašius, 2014).Furthermore, the first countries to have initiatives under this theme were Germany (2010), the Netherlands (2011), France (2011) and Spain (2011).Nevertheless, every member of the European Union is represented.The last countries to present results were Bulgaria (2017), Luxembourg (2017), Slovakia (2016), Latvia (2016), Hungary (2016), Cyprus (2016), and Croatia (2016).Smart Cities 2022, 5, FOR PEER REVIEW 8

Figure 4 .
Figure 4. Year of existing studies in the literature by country.

Figure 5 .
Figure 5. Clusters of the countries by their number of literature results.

Figure 5 .
Figure 5. Clusters of the countries by their number of literature results.

Figure 6 .
Figure 6.Distribution of the smart city initiative results by category.

Figure 6 .
Figure 6.Distribution of the smart city initiative results by category.

Figure 7 .
Figure 7. Countries' Smart City initiatives divided by category.

Figure 7 .
Figure 7. Countries' Smart City initiatives divided by category.

Figure 8 .
Figure 8. Categories with the widest presence among countries.

Figure 8 .
Figure 8. Categories with the widest presence among countries.

Figure 9 .
Figure 9. Geographic representation of the scores of the factors.

Figure 9 .
Figure 9. Geographic representation of the scores of the factors.

Figure 10 .
Figure 10.Loadings of the principal component analysis.

Figure 10 .
Figure 10.Loadings of the principal component analysis.

Figure A2 .
Figure A2.Geographic representation of the scores of the factors.

Figure A3 .
Figure A3.Cluster analysis of the countries by the percentage of the initiatives.

Table 2 .
Loadings of the principal component analysis.

Table A5 .
Socio-economic characteristics of clusters.