A Network Approach to Spatial Data Infrastructure Applying Social Network Analysis in SDI research

Transcription

1 A Network Approach to Spatial Data Infrastructure Applying Social Network Analysis in SDI research Glenn Vancauwenberghe, Geert Bouckaert and Joep Crompvoets K.U. Leuven, Public Management Institute, Abstract So far, most SDI-research is descriptive in nature. In order to tackle the weaknesses of the existing SDI descriptions, this paper introduces a network approach for SDI. However, specific models and methods are needed to analyze SDI from this network perspective. This paper examines how these models and methods can be provided by (social) network analysis. Therefore, this methodology is applied in the context of the Flemish SDI. Keywords: spatial data infrastructures, research methodology, network approach, network analysis 1. INTRODUCTION At different levels of society spatial data infrastructures (SDI) are developed to facilitate and coordinate the exchange and sharing of spatial data (Crompvoets, et al, 2004). Describing the development of these infrastructures forms a substantial part of current research on spatial data infrastructures. Within these descriptive studies, SDI initiatives at several administrative levels are analyzed and compared (e.g.: Vandenbroucke et al, 2008; Masser, 2005; Craglia et al, 2003). The so-called SDI components play an important role in these descriptions. As SDI s are often defined in terms of the components necessary to develop them, these components are easily used to determine the categories the comparisons are based on. This component-based view on SDI and SDI development is clearly expressed in many SDI description studies (REF). However, few attention is paid to the actual results of spatial data infrastructures, that is to say the effect on practices of data acquisition and data sharing. The question to what extent the access to and the use of spatial data are improved due to the SDI initiatives remains unanswered. Particularly from an evaluative point of view, this might be dangerous. The fact that certain SDI components are well developed says nothing about their contribution to the sharing of spatial data. In addition, the majority of the descriptive studies offers a description of single SDI initiatives, or initiatives at a single administrative level. In these studies, a prominent place is given to initiatives at federal level. SDI initiatives at lower levels of administration are recognized, but the attention they receive is remarkably limited (Craglia et al, 2004). Nevertheless, initiatives at local, provincial or regional levels can be decisive for the access, exchange and use of spatial data as well. Moreover, the use of spatial data by a public organisation will rarely be determined by a single initiative. In reality public organisations are confronted with a collection of SDI initiatives. The issue of how these initiatives are related to each other has always been an important subject within the field of SDI research. Research to date has tended to start from a hierarchical relation between SDI initiatives at different levels (Rajabifard et al, 2003). Although some efforts at creating SDIs at different levels strongly follow this hierarchical model, more complex interactions can also be detected (Budhathoki et al, 2007). In order to describe and analyze these complex interactions new models and methodologies are needed.

2 Describing SDI initiatives and their underlying components unarguably has a high value in current SDI research. The limited attention that is given to the impact of these infrastructures imposes a fundamental restriction. The presence of a spatial data infrastructure does not automatically imply that (all) public organisations make use of this infrastructure to acquire the spatial data they need. For various reasons of technological and non-technological nature organisations might prefer other options. At the moment, sharing and exchanging is just one alternative among many others to acquire spatial data. For example, despite the presence of accessible and usable geographic data, some organisations still prefer to produce their own data, with or without the help of private organisations. Discovering the reasons behind this situation is an important challenge for SDI research. Summarizing, the strong emphasis on single SDI initiatives, and - within this initiatives on SDI components, is one of the weaknesses in traditional SDI description studies. Through analyzing these initiatives from a more organisational point of view, we will tackle this problem. Additional to traditional approaches for defining and operationalising SDIs, we therefore propose a network approach on SDI (Vandenbroucke et al, 2009). In this approach, an SDI is operationalised in terms of the organisations that are producing, using and sharing spatial data, and in terms of the flows of spatial data between these organisations. Together they form a network of spatial data sharing: the producers and users of spatial data are the so-called actors in this network, the flows of spatial data between these actors give shape to their relations. A spatial data infrastructure can then be seen as the collection of arrangements determining or influencing these organisations and their relations. In that way, the organisations, data flows, and infrastructure form three separate dimensions. The network approach on SDI strongly emphasises that all three dimensions are crucial in SDI development. Research in the field of SDI can contribute to this development by providing separately insight to each of these dimensions. However, more critical is the interaction between these three dimensions. The strength of our network approach is that it provides a framework to integrate the different dimensions of SDI and SDI research. Research on SDI initiatives (and their components) can thus be complemented with analysis of how these initiatives give shape to the exchange of data between organisations and the use of data within organisations. The aim of this paper is to demonstrate the value of this network approach for research in the field of SDI development. To analyze SDI from this network perspective specific models and methods are necessary. This paper examines how these models and methods can be provided by network analysis. We therefore apply this methodology on practices of spatial data exchange in the region of Flanders 2. NETWORK ANALYSIS Network analysis is often introduced as an alternative to the traditional research approaches in social sciences (Scott, 2004; Knoke and Yang, 2008; Wasserman & Faust, 2008). By comparing social network analysis with these individualistic and more variable-based approaches, the strengths of this alternative method are underlined. Knoke and Yang (2008) point out that in social sciences the assumption is traditionally made that individual actors act and make decisions without regard to other actors. Explanations of this decisions and actions are then mainly based on the individual characteristics of these actors, and less on the external setting. Network analysis starts from the statement that actors are part of social systems connecting them to other actors. In determining the behaviour of actors, their structural relations with other actors are more significant than individual characteristics. These relations, and their form and content in particular, are the main research object in network analyses.

3 The difference between both research traditions can also be clarified by having a look at the empirical data their working with. In general, two types of data can be discerned: attribute data and relational data (Scott, 2004). Attribute data refer to the personal characteristics of individuals, organisations or groups. This type of data is often used to explain certain attitudes or behaviour, the so-called variable analyses. Relational data concern contacts of connections between individuals or groups. These relations then are not properties of these individuals, but properties of the whole system of individuals. They connect these individuals into a larger relational system. Social network analysis offers the methods to analyze these relations. Both types of data thus demand a specific analytical approach. For what concerns the collection of these data, no distinction has to be made. Just as attribute data, relational data can be collected through several methods. Although the use of questionnaires is the most appropriate method, interviews, observations and secondary resources can as well be employed in social network analyses. In general terms the concept of a network refers to patterned relationships among individuals, groups or organisations (Dubini et al, 1991). Any network thus encompasses two indispensible elements: actors and relations (Knoke et al, 2008). These actors can be all types of social entities: individuals, groups, organisations, nation-states, etc. (Wasserman and Faust, 2008). Actors are linked to one another by relational ties. These relational ties too can be of various nature. Affective relations, exchange of goods or information and personal contacts are examples of common types of relations (Knoke et al, 1982). Network analysis can then be defined as an empirical tool to describe, measure and analyse social structure on the basis of the multiple sets of relations between these actors (Keast et al, 2005). Within network analyses the emphasis lies on relations between a certain set of actors. Two properties of these relations are relevant to further analysing them: whether they are directed or non-directed, and whether they are dichotomous or valued (Wasserman et al, 2008). In a direct relation the tie starts form one actor and goes to the other, while in non-directed relations two actors are connect to each other, without a certain direction. In addition, a relation can be dichotomous or valued. In dichotomous relations, information is only collected on the presence or absence of a tie between two actors. In case of valued relations ties are further specified with a value, as an indication for the strength or frequency of the relation. In most network analyses only one type of relationships, possibly directed or value, is observed. However, it is possible to include and analyse several types of relationships into one network. For example, private organisations can exchange both information and financial resources, or students can be both friends and classmates. At this moment, methodologies to analyse these multi-relational data are less developed than those for handling single relations (Hanneman et al, 2005). Although social network analysis places emphasis on data of relational nature, the methodology also offers the opportunity to include information on the characteristics of the actors in the analyses. It is argued that these characteristics can influence the structure of the network (Wasserman et al, 2008). These attribute data are exactly the same data that are used in non-relational research methods. Examples of these are age and sex in case of individuals, or location and size in case of organisations. Two sets of analysis can be distinguished within network analysis as analytical tool: graphical and mathematical (Keast et al, 2005). Graphical analyses concern the visual representation of networks. These visualisations are based on so-called graphs, in which actors

4 are represented as points or dots, and relations are expressed through lines connecting these points. Graphs are praised for their strong exploratory and explanatory power because they presented networks and their relations in a clear and concise way (Brandes, et al, 2006). Mathematical analyses are more advanced, here calculations and statistical analyses are performed on the collected data. In order to perform these analyses, all relational data need to be represented in a socio-matrix. In this matrix all actors are listed in one row and one column, and information about the relational ties among the actors is displayed in the cells of the matrix. Compared to graphs these socio-matrices are notably less user-friendly, but offer the opportunity for in-depth analysis (Knoke et al, 2008). Furthermore, analyses can be performed at several levels, using specific methods and measures. Typical levels of analysis are actors, subsets of actors and the entire network (Wasserman et al, 2008). 3. THE USE OF NETWORK ANALYSIS IN THE SPATIALIST RESEARCH PROJECT In September 2007, Katholieke Universiteit Leuven and Vrije Universiteit Brussel started a four-year research project entitled SPATIALIST; Spatial Data Infrastructures and Public Sector Innovation funded by Institute for the Promotion of Innovation by Science and Technology in Flanders (Crompvoets et al, 2008). The research object of the project is the relation between Spatial Data Infrastructures (SDI) and Public Sector Innovation in the Flemish Region. The SPATIALIST research project aims to identify the key requirements to further develop, operate, and maintain a successful Flemish SDI. The project started with a so called zero measurement that resulted in a general overview on how spatial data and GI-technology is currently tied up with organisations in the Flemish public sector. Data were collected through an online questionnaire, sent to all public organisations in Flanders. In total, 508 public organisations in Flanders were invited to participate in the survey. Private organisations and non-profit organisations were left out of the research population, as the SPATIALIST-project focuses on SDI in the public sector. The SPATIALIST-questionnaire was completed by 234 public organisations in Flanders, at different administrative levels. These organisations constitute the actors in our network. A first set of questions the survey was designed to identify some of the characteristics of the organisations. Information was collected on their administrative level, their use of GIS, their data policy and their organisational structure. This information can be entered in the network analysis, in the form of attribute data, which offers additional options for representing and analysing our network. As repeatedly mentioned before, network analysis mainly focuses on the relations among these actors. In the context of spatial data infrastructures and spatial data sharing, relations of data exchange are most relevant. Therefore, the SPATIALIST-survey was largely directed at identifying flows of spatial data among public organisations in Flanders. In order to identify these data flows in a clear and unambiguous manner, four specific types of geographic data were selected: parcel data, address data, road data and hydrographic data. For each of these four types of spatial data, the organisations in our population were asked to indicate whether they used, produced and/or received these data. In case they received a certain type of data from one or more other organisations, they were requested to identify the supplier(s). To this end, we made use of a so-called fixed choice roster. This means all suppliers had to be selected from a list of actors, but the number of choices was restricted to five. Like that, we were able to construct one multi-relational network of spatial data flows in Flanders or four separate networks, as information was collected on four different types of spatial data.

5 In a next section of the survey, questions were asked on the characteristics of these data flows. Some examples of this are the price, the legal basis, and the type and frequency of transfer. These characteristics offer additional insight in the observed data flows. In network analyses, they can be applied in several ways. For example, they can be used to valuate relations. From the perspective of optimal spatial data sharing, one could rate free data flows more positively than charged flows. Furthermore, these data flow characteristics are also usable to distinguish different networks. In that case, one could compare a free network with a charged one. The SPATIALIST zero measurement demonstrates how a network approach on SDI offers an interesting perspective for designing SDI research. With some simple examples, we will now illustrate the analytical value of this network approach and the related network analysis. Earlier, we mentioned that analyses were conceivable at different levels. Moreover, at each level specific models and methods could be used. Concerning analyses of the complete network, a frequently used measure is the density of the network (Wasserman & Faust, 2008). Density refers to the degree to which actors are connected. As a measurement it can be seen as the ratio of existing connections to all possible connections. The higher this ratio, them more denser the network is. In SDI research this measure can be used to compare the exchange of different types of spatial data. Table 1 presents the density of the four networks of geographic data that were analyzed in the SPATIALIST- zero measurement. It can be seen from the data presented in this table that the parcel data network is the most dense. This means that this type of spatial data is exchanged most frequently. However, these results must be interpreted with caution. After all, the development and implementation of spatial data infrastructures is not aimed at exchanging spatial data as much as possible. Frequent exchange does not automatically imply optimal exchange. Future studies on the exact interpretation of density in the context of spatial data infrastructures therefore are recommended. Table 1: Density of the four networks in the SPATIALIST-zero measurement (preliminary results) Density Parcel data 0,0119 Address data 0,0062 Road data 0,0046 Hydrographic data 0,0044 At the actor level, the centrality measure provides insight in the involvement of actors in all relations within a network. Centrality refers to the degree to which network activity is centred on one or a few actors (Keast et al, 2005). The relevance of this measure in SDI research is shown in table 2. This table presents the four most central actors in the parcel data network in Flanders. It is apparent from this table that for what concerns the exchange of parcel data, there are two key actors: the Federal Public Service Finance and the Agency for Geographic Information Flanders. Further research should be done to investigate the relation between both organisations. The results in table 2 do not show whether these two central organisations overlap or complement each other. They neither provide insight in the implications of this centralized management of parcel data in Flanders.

6 Table 2: Centrality in the parcel network (preliminary results) Centrality Federal Public Service Finance 89,000 Agency for Geographic Information Flanders 62,000 Flanders Oost Vlaanderen 17,000 R-O Vlaanderen 7,000 One of the most interesting features of network analysis is the potential to visualize networks in a direct and incisive way. The graphical representation of networks can contribute to the creation of new insights and knowledge (Brandes et al, 1999). To conclude this section of the paper, we illustrate this with an example. Figure 1 gives an overview of the flows of parcel data going from the five provinces in Flanders to the local governments. From this figure we can see the differences in spatial data sharing in the five Flemish provinces. While in the province of Oost-Vlaanderen the majority of the local administrations obtains parcel data from the provincial administration, this is not the case for the other provinces. Particularly in the province of Antwerpen, the flow of parcel data between provincial administration and the local authorities is extremely limited. This figure provides an interesting starting point for further research. A first research question that could be asked concerns the explanation of current situation. More specific we need to analyze the role of both administrative levels in shaping this reality. On the other hand future research is required on the exact implications of this situation. Figure 1: Provincial-local relations in the parcel network (preliminary results)

7 4. CONCLUSION: THE INTEGRATIVE VALUE OF NETWORKS The ambition of this paper was to introduce our network approach as a distinct strategy to describe and analyse spatial data infrastructures. Network analysis can then be seen as a methodological tool to put this network approach into practice. To conclude this paper, we briefly discuss some of the strengths of both the network approach in general and network analysis in particular. These strengths all concern the integration of traditionally separated or underrepresented aspects of SDI research. 1. Integrating organisations, data flows and SDI As mentioned at the beginning of this paper, traditional SDI research places strong emphasis on SDI initiatives and their components. As a result, few attention is paid to the results of these initiatives, namely their actual impact on the exchange and use of spatial data. With our introduction of the network approach we try to tackle this problem. In addition to the infrastructure and its initiatives, the network approach looks at all organisations using this infrastructure and the resulting flows of spatial data. So, the first strength of this approach is the integration of the different levels into one framework. 2. Integrating SDI initiatives Moreover, SDI description studies traditionally focus on one or few SDI-initiatives in a certain region. Hence the assumption is made that all initiatives are linked to each other in an hierarchical way. In reality, however, organisations are confronted with a series of initiatives, often associated to a certain policy domain or a certain administrative level. Moreover, relations between these initiatives of different kind, are rather complex. Consequently, an hierarchical view on SDI is not always tenable. New methods and models are demanded to grasp these complex reality. The network approach on SDI offers a way out, because she takes organisations producing and using spatial data as a starting point. Analysing SDI in terms of these organisations, can help us to gain more insight in the complex relations between different SDI initiatives. 3. Integrating methods of data acquisition Spatial data infrastructures are developed to facilitate and coordinate the exchange of spatial data. So it can be argued that these infrastructures aim to effect the way organisations acquire their spatial data. The extent to which this aim is reached demands special attention in SDI research. In order to meet this demand, more insight is needed in the different manners of acquiring spatial data. In general, two methods of data acquisition can be discerned: producing new spatial data or using existing data. Within both options, once more several alternatives can be distinguished. In order to assess the performance of a spatial data infrastructure, it is essential to map these different options. Again, our network approach offers a suitable method to do this. 4. Integrating research disciplines Finally, as the development of spatial data infrastructures will have to meet requirements of different nature, it is one of the challenges in current SDI research to handle these requirements. The SPATIALIST project takes up this challenge through a multi-disciplinary research design. In this design, the traditional technological requirements of SDI are supplemented with legal, economic, organisational and interorganisational requirements. The network approach on SDI fits in with this multi-disciplinary character of SDI, as each of these disciplines can be operationalised on our three dimensions, namely the organisations, data flows and the SDI initiatives. For example, the economic research questions of SPATIALIST will simultaneously

8 look at the financial mechanisms within an organisation, the financing of a specific data flow and at agreements on funding and pricing within a certain SDI initiative. REFERENCES Brandes, U., Kenis, P., Raab, J., Schneider, V. and D. Wagner (1999). Explorations into the visualization of policy networks, Journal of Theoretical Politics, 11(1): Brandes, U., Kenis, P. and J. Raab (2006). Explanation Through Network Visualization, Methodology, 2(1): Budhathoki, N. and Z. Nedovic-Budic (2007). Expanding the spatial data infrastructure knowledge base, in Onsrud H. (Ed.). Research and Theory in Advancing Spatial Data Infrastructure Concepts. Redlands, CA: ESRI Press, pp Craglia M. et al. (Ed.) (2003). GI in the Wider Europe, at [accessed 14 Februari 2009]. Craglia, M. and A. Johnston (2004). Assessing the Impacts of Spatial Data Infrastructures: Methods and Gaps, 7th AGILE Conference on Geographic Information Science, Heraklion, Greece. Crompvoets, J., Bregt, A., Rajabifard, A. and I.,Williamson (2004). Assessing the worldwide developments of national spatial data clearinghouses, International Journal of Geographical Information Science, 18(7): Crompvoets, J., Bouckaert, G., Vancauwenberghe, G., Vandenbroucke, D., Van Orshoven, J., Janssen, K. et al. (2008). Interdisciplinary research project: SPATIALIST. Spatial Data Infrastructures and Public Sector Innovation in Flanders (Belgium), GSDI-10 Conference, Trinidad: St. Augustine. Dubini, P. and H. Aldrich (1991). Personal and Extended Networks Are Central to the Entrepreneurial Process, Journal of Business Venturing, 6(5): Freeman, L. (2000) Visualizing Social Networks, Journal of Social Structure, 1(1). Hanneman, R. A. and M. Riddle (2005). Introduction to Social Network Methods. Riverside, CA: University of California, at [accessed 7 August 2008]. Keast, R. and K. Brown (2005). The Network Approach to Evaluation: Uncovering Patterns, Possibilities and Pitfalls, Australasian Evaluation Society International Conference. South Bank, Brisbane. Knoke, D. and J. H. Kuklinski (1982). Network Analysis. Newbury Park: Sage. Knoke, D. and S. Yang (2008). Social Network Analysis. Los Angeles: Sage Publications. Masser, I. (2005). GIS Worlds Creating Spatial Data Infrastructures. Redlands, CA: ESRI Press. Rajabifard, A., Feeney, M.-E. F. and I.P. Williamson (2003). Spatial Data Infrastructures: concept, nature and SDI hierarchy, in Williamson, I.P., Rajabifard, A. and M.-E. F. Feeney (Eds.). Developing Spatial Data Infrastructures. From concept to reality London / New York: Taylor and Francis, pp

9 Scott, J. (2000). Social Network Analysis: A Handbook. London: Sage Publications. Vandenbroucke, D. and K. Janssen (2008). Spatial Data Infrastructures in Europe: State of Play Leuven: Belgium. Vandenbroucke, D., Crompvoets, J., Vancauwenberghe, G., Dessers, E. and J. Van Orshoven (2009). A network perspective on Spatial Data Infrastructures: Application to the subnational SDI of Flanders (Belgium), Transactions in GIS [under review]. Wasserman, S. and K. Faust (2008). Social Network Analysis: Methods and Applications. Cambridge: Cambridge University Press.