Exploring Stockholm’s Spatial Capital in Relation to Sustainable Development:

a quantitative study on the impact of geographical proximity on education and income

Tobias Karström

June 2018 Supervisor: Michael Meinild Nielsen Department of Human Geography Stockholm University SE-106 91 Stockholm / Karström, Tobias. 2018: Exploring Stockolm’s Spatial Capital in Relation to Sustainable Development: a quantitative study on the impact of geographical proximity on education and income.

Advanced level: Master thesis Master’s degree in Urban and Regional Planning, 30 ECTS Supervisor: Michael Meinild Nielsen Language: English

ABSTRACT

The aim of this thesis was to examine if the theoretical concept of spatial capital explains the distribution of the human capital and high incomes in an urban region using Stockholm county as an empirical example. The spatial capital theory suggests that geographical proximities can explain social and economic distributions in urban regions and is divided into two sub- categories; position and situation capital where the former measures the geographical proximities to urban environments and cultural amenities and the latter measures to proximities to public transportation alternatives. The study was conducted using proximity analysis performed with Geographical Information Systems (GIS) in QGIS, and stepwise multiple linear regression analysis performed in SPSS. Human capital and high incomes were chosen as dependent variables due to their relevance to sustainable development, especially in regards to urban economics becoming more knowledge-based. Spatial capital was operationalized into five variables of measuring geographical proximities; distance to central business district (CBD), distance to regional urban cores, distance to nearest amenity, distance to nearest railway, and distance to nearest bus stop. The human capital was operationalized as tertiary education and income was measured as taxable earned income. All results demonstrated statistically significant effects on the dependent variables. Proximity to central Stockholm was the strongest explanatory variable regarding the distribution of human capital and high incomes in Stockholm county. When excluding this variable, distance to railways was proven to have the strongest effect on the distributions. The conclusion was that the spatial capital theory explains that the distance to central Stockholm and distance railways are the strong predictors for how human capital and high incomes are distributed in the county. The robustness of analysis and the empirical findings’ meaning in the context of regional planning in Sweden was then presented and discussed.

Key words: Spatial capital, human capital, regional planning, Stockholm county, sustainable development

2 Acknowledgement

Writing this master’s thesis has been as enjoyable as frustrating and has cost me many hours of writing, rewriting, erasing, editing, reediting, and more writing. It has also cost me countless amounts of coffee to keep myself going, and a glass of wine every now and then to help me unwind.

I would like to express my deepest gratitude towards my supervisor Michael M. Nielsen for all his support and constructive feedback, and also for introducing me to the art of using geographical information systems. I would also like to thank my friends and colleagues for all their advices and all their support during this spring. And finally I would like to thank Sofia, my dearest, for everything.

Tobias Karström Tullinge, May 2018

3 ABSTRACT ...... 2 Acknowledgement ...... 3 1.0. Introduction ...... 5 1.1. Aim of thesis and research questions ...... 6 1.2. Disposition ...... 6 2.0. Background ...... 7 2.1. Part I: The Swedish planning process and urban sustainability ...... 7 2.1.1. The planning process ...... 7 2.1.2. Regional planning in Sweden ...... 7 2.1.3. Contemporary issues of social segregation ...... 9 2.2. Part II: Exploring the origin of the spatial capital theory, its background and application...... 10 2.2.1. An historical overview ...... 10 2.2.2. The Social Logic of Space, and Space Syntax modelling ...... 11 3.0. Theoretical framework: Spatial capital and human capital ...... 13 3.1. Part I: The theoretical issue of spatial capital, the meaning and the measure in the context of urban and regional planning ...... 13 3.1.1. Spatial capital as a value capture ...... 14 3.1.2. Spatial capital as a time-geographical concept ...... 14 3.1.3. Spatial capital as a tool for urban planning ...... 15 3.1.4. Spatial capital as a theory on gentrification, , and the rise of the creative class ...... 17 3.2. Part II: Social sustainability and its relationship to spatial capital ...... 18 3.2.1. What is sustainable development, and social sustainability in particular? ...... 19 3.2.2. The link between social and human capital ...... 20 3.3. Theoretical conclusion: The geographical link between urban sustainability ...... 21 4.0. Methodology: quantitative design and hypothesis testing ...... 22 4.1. Research design ...... 22 4.1.1. Operationalization of theory and data used ...... 22 4.1.2. Research object and cartographical data ...... 24 4.2. Execution ...... 25 4.2.1. Geographical proximity analysis ...... 26 4.2.2. Stepwise multiple linear regression analysis ...... 26 4.2.3. Key statistical concepts and meanings ...... 27 4.3. Hypothesis testing and methodological strengths and limitations ...... 27 4.3.1. Hypotheses ...... 28 4.3.2. Ethical considerations and evaluation of sources ...... 29 5.0. Results ...... 30 5.1. Proximity analysis and descriptive statistics ...... 30 5.2. Standardized residuals ...... 31 5.3. Summary of coefficients ...... 34 5.4. Collinearity diagnostics and auto-correlation ...... 36 5.5. Exclusion of X1: Distance to the CBD ...... 36 5.5. Maps over the residuals’ spatial distribution ...... 38 5.6. Hypothesis testing ...... 39 5.7. Results summary ...... 41 6.0. Discussion ...... 41 6.1. Discussion on results and methodology ...... 41 6.2. Exploring Stockholm’s spatial capital ...... 43 7.0. Conclusion ...... 44 References ...... 46 APPENDIX ...... 52

4 1.0. Introduction

The aspect of land and how to use and access it is of great importance to urban and regional planning. Historically, the implementation of transportation systems and infrastructural constructions has increased the value of land as a result of enabling transportation alternatives. For instance land-owners in Illinois, United States experienced significant capital gains in the mid 1800’s due to the construction of railways that were located in close proximity to their land, causing increasing market values (Coffman & Gregson, 1998). After decades of rapid urbanization, the concept of sustainable development has finally gone mainstream – in short this means that urban planning emphasises to untie tensions between economic and social development while simultaneously prevent further damage on the Earth’s biological ecosystems (Gunnarsson-Östling et al., 2013: p. 53). For this reason, the intellectual conceptualization of sustainability as a political as well as a scientific discourse is a complex topic. Western metropolitan cities have become the ‘experimental platforms’ of sustainable living and are experiencing increased living expenses in urban environments that are optimal for these lifestyles that sets the political and economical standard for how urban planning is conducted (ibid). In the case of Stockholm, Sweden which was branded as ‘greenest capital’ by the European Union in 2010, the city can ben seen as an ideal-type for what a ‘sustainable’ city is according to international standards (Metzger & Olsson, 2013: p. 1). It raises a question whether the spatial elements of the city can explain the social and economic mobility of its inhabitants and thus how urban sustainability is achieved.

For this reason, this thesis focuses on a theoretical concept referred to as spatial capital which examines the effects of proximities. The concept can be studied in a variety of ways and has for instance been used to analyse how street patterns in cities can inform of fundamental values represented by land (see Marcus, 2008, 2017) and how it interconnects with aspects such as infrastructure, buildings and roads and spatial forms that are prominent for the . The values represented by land concerns both use-values for humans in their everyday-life, as well as the market values represented by different costs. The spatial capital concept from this point of view argues that these are correlating phenomena, where the quality of the land itself is the predicting variable. However, geographical knowledge does not only concern the physical aspects of space, but equally emphasises the timely elements of the human perception on space – mobility and accessibility are two components that are of high value when planning for transportations and for placements of schools and other public domains. The spatial capital concept has therefore also been used in the study of geographical proximities and the travel- time for different demographics in cities (see Lévy & Lussault, 2003; and Barthon & Monfroy, 2010). Their perspectives suggest that urban planning can use the timely elements to compensate for the physical elements by enriching the mobility in cities, thus ‘minimizing’ distances. The spatial capital concept clearly emphasises the distances to places as well the places themselves, and what values they contain (ibid). How is spatial capital then relevant for sustainable development?

A popular idea about the ‘sustainable city’ is that it requires a knowledge-based economy where the growth is largely generated by post-industrial or non-material industries pioneered through technological solutions and digitalisation (see Florida, 2002, 2006; and Glaeser et al., 1992, 2004). The labour market in fields such as IT, engineering, public policy et cetera. requires specialisation most likely achieved through tertiary education – when cities become more specialised, the competition for optimal spaces could increase which is reflected in the costs for land-use, the housing and office rents (see Marcus, 2017). The enablement of ‘sustainable’ lifestyles is then dependent on individuals having the accessibility to this social arena, or the

5 creative class as Richard Florida (2002) understands it. The more established theoretical concepts social capital which studies individuals’ collected social networks and ties (see Putnam, 1994) and human capital which studies individuals’ collected competencies and talents (see Becker, 1964) are two central concepts in how the ‘sustainable city’ is achieved (Marcus, 2010; Liljefors, 2013, and Weingaertner & Moberg, 2014). The question that this thesis examines, is whether the spatial capital, understood here as the collected geographical proximities of places can explain how education levels and individuals with high incomes are distributed in Stockholm county and what this means for sustainable development.

1.1. Aim of thesis and research questions

The aim of the master thesis is to examine if the theoretical concept of spatial capital explains the distribution of the human capital and high incomes in an urban region using Stockholm county as an empirical example. The aim is also to analyse how these distributions relates to theories on urban sustainability, with a focus on social sustainability in particular. The reason for choosing Stockholm as the object of study is that the city has been internationally recognized as an ideal-type of a sustainable city (Metzger & Olsson, 2013). It is also the thesis’ ambition to discuss whether the theoretical concept of spatial capital can be used as a synergetic framework for analysing social, cultural and economic patterns in relation to the empirical findings, with a focus on the regional planning in Sweden.

The research question that the thesis aims to answer are:

I. Are the distances to central Stockholm, the regional urban centres, cultural amenities as well as public transportation correlating with the distribution of the human capital as well as the incomes of individuals living in Stockholm county? II. Are the spatial distributions of human capital and incomes recognized or critically discussed in the theoretical understanding for social sustainability? If so, how does it relate to the regional planning process for Stockholm county?

1.2. Disposition

The thesis is divided into seven chapters. The first one being the introduction which presents the thesis’ topic, aim and research questions. The Second chapter introduces the background to the Swedish planning process with a focus on the political aspects and potential discrepancies between strategies for planning on macro vs. micro levels in regards to sustainability, and social segregation in Sweden. It also introduces the origins of the spatial capital concept. The third chapter introduces the theoretical framework of spatial capital and the connection to human capital and income that the study is based on. The fourth chapter presents the research methodology and the empirical design of the study. The fifth chapter presents the results and an analysis of the findings by presenting maps, tables and diagrams. The sixth chapter presents a critical discussion on the results and the methodology being used, as well as a discussion on the subject matter and recommendations for future research. Chapter eight concludes the thesis’ findings, answers the research questions and evaluates the study’s contributions to the research field. Finally, a reference list of the literature and an appendix explaining the empirical material is placed at the end of the thesis.

6 2.0. Background

In order to answer the research questions, some basic knowledge about the Swedish planning process is in order. This chapter therefore presents how urban and regional planning is practised in Sweden and focuses on the history and the development of the current political outline and the legislative actions. It introduces Sweden’s approach to regional planning in particular, and the goals for achieving urban sustainability in the Stockholm region. The chapter also introduces contemporary socio-political issues of social segregation in Sweden, and how it relates to urban planning throughout the 20th century as well as an introduction theoretical origins to the spatial capital theory and how it subsequently has been elaborated in in different fields.

2.1. Part I: The Swedish planning process and urban sustainability

2.1.1. The planning process The foundations of the Swedish urban planning process stretches several centuries back, but the overall principles were formulated during the 1900’s as a result of the rapid urbanization and population boom, as well as the construction and the development of the welfare state. During the early years, City plans required authorization by the King of Sweden, but due to political reformations that emphasised the need for cheap housing and greater social cares, municipalities were provided with the authority to regulate and exploit their own lands. The purpose was that municipalities would a be able to influence housing policies and find the optimal spaces for housing locations (Boverket, 2014). The political principle of local governance laid the grounds for the municipal planning monopoly in 1947 which was strengthen in 1960 due to discrepancies in local building strategies. This resulted in a new legislation where municipalities were obligated to implement a local council that overlooked the building process by state-regulated means and provide with more standardized building permits for the entire nation, even if the municipalities still remained responsible for the exploitation (ibid). In 1987 the Planning and Building Act (Swedish: Plan- och bygglagen, PBL) was implemented as a result of political failures to coordinate interests that was largely caused by the conflicts between the many laws that regulated the use of land and water. The motivation for implementing PBL was to find a more holistic approach that would clarify different sectors’ political responsibilities and stimulate a democratic planning process where local citizens’ interests were more considered than before. The main principle of PBL was the reformation of the planning process by separating between the comprehensive planning (Översiktsplanering) and the detailed zoning (Detaljplanering). Every municipality is now obligated to produce a comprehensive plan that will cover all land and water supplies within the geographical boundaries and provide with alternatives for future development in regards to the national political ambitions. The detailed zoning on the other hand covers a smaller geographical area that foundations for the legislation of the use of land that in turn needs to be approved by the state. In 2011 PBL was reformed which provided with a stronger focus on the comprehensive planning, especially in regards to environmental factors such as climate change, and the development of electronic communications (ibid).

2.1.2. Regional planning in Sweden As emphasised in the previous section, municipalities have profound influences on the planning process, enabled by the monopoly for land regulations. Regional planning, which can be understood as the planning for a larger geographical context including many municipalities, has

7 a much more limited influence and functions more as a development framework (Boverket, 2015) Its influence is also depending on whether a region mostly covers rural and urban municipalities. In the latter case, physical development such as dwellings and infrastructure are more crucial questions in regards to economic growth and comes with increasing housing and transportation demands (Boverket, 2015; and 2017). In the case of Stockholm county which consists of 26 different municipalities, regional planning is largely a question of utilising the strategies for maintaining the county’s economic growth as well as influence the local strategies for maintaining sustainable development like social equity and reducing CO2-emmisions (Liljefors, 2013). In order to ‘build bridges’ between urban geographies, the concept of regional planning is in theory related to concept of polycentrism (see Schmidt et al., 2006) which in the case of Stockholm county would mean the interaction and connectivity in and between the regional urban cores (see Chapter 4 p.26 and Map 2) and the central city area. A regional urban core within the county (for instance Skärholmen-Kungens Kurva) can be understood as an example of a polycentric node that crosses municipal boundaries1 within a geographical area and is defined by (for most part) entrepreneurship and commercialism. In order for these areas to flourish, their accessibility becomes of great importance. However, one could easily imagine that political conflicts arise from the planning process when interests are cross-sectorial and when the macroscopic visions provided by the regional planning meets the legislative organs on the micro level.

In the aspect of planning in Sweden, the political frictions between geographical scales has been studied by Blücher & Graninger (2006) who express that post-1970’s political forces have questioned whether the municipal planning monopoly should be abolished as its original intention of providing inhabitants of cheap housing has somewhat changed – arguing that municipalities rather regard their land as a commodity that provides with high investment returns when selling pieces of the land to real-estate owners. Thus the monopoly makes the demand for land quite limited and exclusive, resulting in heavy costs for exploitation and building-permits which is making high standard housing that are affordable for the majority of the population very difficult. The arguments for abolishment of the monopoly according to Blücher & Graninger implies that the increase of land supplies would press down the prices (ibid). From to perception of regional planning, the principle of local governance becomes a valid starting point for critically exanimating what role the regional framework has on spatial planning practices, especially in regards to sustainability approaches that are macroscopic and very general in their formulations, and the question of where and how to implement them. For instance, Stockholm county has produced a regional development plan called Regional utvecklingsplan för Stockholmsregionen (RUFS) which expresses goals set up by the city in order to achieve urban sustainability till the year 2050. These includes, amongst others, following ambitions:

“To reduce climate impact and at the same time enable greater accessibility and economic growth” (RUFS, 2016: p. 5).

”[Identifying] how the physical planning can create better conditions for social cohesion in the region and contribute to an equal society” (ibid: p. 9)

The plan states that one of the goals for improving connectivity in Stockholm in order to promote the regional growth. It also emphasises how physical planning is relevant to promote

1 Skärholmen is located in Stockholms stad and Kungens kurva in Huddinge.

8 social cohesions and contribute to social equity – with the issue of terminating housing shortages and promoting densification of buildings near metro- and train stations being priority goals, as well as stimulating a better access to the housing market for individuals who find it difficult to enter it in the current political and economic situation (ibid: p. 9 & 12).

2.1.3. Contemporary issues of social segregation In regards to the Swedish planning process and the sectorial conflict between regional planning and municipal governance stated in the previous section, the ambivalence of power in the planning process becomes a relevant topic. According to Friedmann (1998) urban planning is not a process that requires a de facto rational design applied to all cities. Historical contexts and spatial qualities are contributing factors to how cities develop and progress. In the case of Stockholm, the issue of social segregation is prominent, which usually goes in hand with spatial distributions – differences in population patterns in cities and regions often correlate with different opportunities to housing, education and working life which is also known as social segregation (see Andersson & Malmberg, 2016), and results in social inequalities regarding the chances for inhabitants to improve their livelihoods. In this context, segregation can be understood as a synonym to ‘social exclusion’ from spatial contexts with ’better’ living conditions, implying that social exclusion as such can be related to physical environments and their spatial interrelations.

The political problem of social segregation in Swedish cities has been studied and analysed for decades and rose to attention during the 1950’ and 60’s, especially after the implementation of Miljonprogrammet (eng: the million programme). This was a political decision to build a million dwellings in a 10-year period stretching from 1965-75 to meet the increasing demands of residence and terminate the issue of housing shortages (Stigendal, 2012). The architectural design was largely characterised by high-rising apartment buildings, often in decentralized locations with poorly developed connectivity and infrastructural solutions for transportation. Even if the constructions were results of a modernist ideal of suburbanization relevant at the time, the areas have suffered a socioeconomic stagnation through the decades and has been perceived as unattractive and would eventually experience a demographic transformation due to population changes and increased immigration (ibid; and Boverket, 2014). Some of these areas have been vulnerable to social anomies such violent crimes and poverty and can be seen as a symbol of social exclusion in Sweden, which would mean that social conditions then has a geographical dimension.

The economic historian and ex-politician for the Swedish liberal party2 Mauricio Rojas authored a pamphlet that illustrated this relationship called Utanförskapets karta (eng: The map of social exclusion), first published in 2004, and expressed that the amount of areas in Sweden categorised as socially excluded3 had gone from 3 in 1990 to 136 in 2002, and to 156 in 2006, which indicates a dramatic increase in a 16 year time period (Sanandaji, 2014: p. 12 ). The Swedish police also identifies certain urban areas where social anomies are a major concern as “particularly socially vulnerable”. As of 2017, these were a total of 23 in the whole of Sweden, most of them are parts of the million programme (Polisen, 2017). The social segregation thus takes an economic, ethnical, and even spatial form. It is evident that the Swedish regional planning for a socially sustainable Stockholm is a complex challenge, and the information

2 Liberalerna; previously known as Folkpartiet liberalerna during this time period. 3 The criteria for social exclusion set by Rojas and later Sanandaji is that at least 60 % of the inhabitants and work-age in the area are unemployed, and that the share of pupils that finish elementary school are lower than 70 %, and (or) the share of voters in previous election was lower than 70 % (Sanandaji, 2014: p. 12).

9 presented in this section suggests that a reduction of social exclusion is necessary for achieving it

2.2. Part II: Exploring the origin of the spatial capital theory, its background and application.

This parts addresses the theoretical development of the spatial capital theory and what intellectual components that has contributed to the concept and how it has been elaborated in previous studies. The section introduces the historical concept that lay the ground for the urban morphological tradition that has developed a theory on spatial capital. The urban morphology tradition is the study of how street network, land lots, building density et cetera. affects human behaviour (Oliveira, 2016) and does not take a stance in this thesis’ theoretical framework. However, the concept does highlight certain aspects of the relationship between humans and the build environment and how the latter can influence the former, and has practitioners has developed a version of approaching a theory on spatial capital.

2.2.1. An historical overview of urban morphological thinking Conzen’s study of Alnwick in Northumberland, in 1960 contributed to the understanding of urban morphology. Conzen’s interest was to solve the problem of how old established town had acquired its geographical complexity and how that knowledge could influence the planning of future towns (Conzen, 1960 in Oliveira, 2016: p. 90). Conzen thereby developed a theory for how to study town-plans by breaking it down to three distinct components of what the urban fabric consisted of: 1) streets and their arrangement in a street system, 2) plots and their aggregation in street blocks, and 3) block plans of buildings (ibid). Conzen would also emphasise the historical dimension of the how these processes can be understood, dividing the physical growth and formation of Alnwick into epochs, or morphological phases and used cartographical elements to analyse how streets, land lots and buildings were spatially connected. This proposed a geographical structure for the study of urban planning as a study of interrelated physical fabrics made by humans with the essential logic being that there is a correlation between historical and social life in how towns are designed and keep on developing (ibid: p. 90f)

Another influential way of study urban morphology is based on the teachings of Kevin Lynch, who popularised the “image of the city” as a methodological approach of explaining the visual experience (or legibility) of a city and how the aspect of urban imagery can help influencing architectural directions (Lynch, 1960 in Oliveira, 2016: p. 92). Lynch based this study on three American cities; Boston, Jersey City and Los Angeles by combing the mapping of what Lynch detected as the five fundamental elements of the urban fabric4 with interviewing citizens of how they visually perceived their cities. Lynch’s conclusion was that an enabled vision of the city is necessary in the finding of a holistic view for how to plan for it (ibid: p. 94). A similar concept was theorized as Townscaping by Gordon Cullen in 1961, who also argued that the urban fabric could be studied from the perspective of the individual’s cognitive experience. Cullen emphasises a study technique based on human movement through urban landscapes in order to detect how textures, colours, styles, and physical layouts of cities can transcend knowledge on its historical development and social dimensions (ibid).

4 According to Lynch these are paths (mainly streets and roads), edges (such as as water drains and train tracks), districts (which shares common characteristics) nodes (Such as junctions and squares), and landmarks (such as distinctive physical objects.

10 A famous contribution from this time period that also came to influence the field of urban morphology, and is widely regarded as a classic with profound influence in urban studies in general is Jane Jacobs’ The Death and Life of Great American Cities from 1961. In it, Jacobs expresses criticism towards the trends of urban planning and architecture that was relevant at the time. Jacobs’ main argument was that the rapid urbanization that took place in America (and in ) during the 1950’s and 60’s which was directly related to population- and economic growth, transformed the way cities were functioning socially. The critique was largely directed towards how the process of urban planning entirely rested on scientific (in this time, positivistic) principles, which didn’t need to emphasise the contexts of how cities work ‘in real life’. One practical issue was how the modern architecture, mostly consisting of high rising buildings alienated people from upholding social control over their territories, or neighbourhoods, which could disable social integration. Metropolitan cities should rather function like smaller cities, which is prevented by large-scale planning practices that deprive urban areas from maintaining their social inclusions and promoting safety. Jacobs argues that in order for planning to achieve social inclusion, neighbourhoods should be rather small-scale, with dwellings being architecturally well integrated with retail and social life, thus strengthening the natural surveillance where people can keep “eyes on the street” (Jacobs, 1961).

In regards to safety, Jacobs’ critical contributions came to inspire the architect Oscar Newman and his theory on defensible spaces. In short, this theory suggests that the architectural design of high rising apartment complexes correlates positively with criminality, with the explanatory factor being the lack of territoriality that the design promotes, making it harder for residents to defend or even ‘care’ for their neighbourhoods (Newman, 1976). Newman based his theory on an observation-study that he conducted in New York City during the early 1970’s and came to be recognized by criminologist C. Ray Jeffery who formulated the concept of Crime Prevention through Environmental Design (CPTED), as a practical framework for architects and city- planners to use in order to design both buildings and whole areas that would effectively decrease the possibility of crime (Ray, 1971).

Even though Conzen, Lynch, Cullen, Jacobs, and Newman may have worked with different strategies and for different purposes, their contributions on the study of urbanity and social life have impacted on the theory of the urban morphology concept and how to understand it. As demonstrated with these theories, the relationship between human action and physical aspects of the built environment is something that has been detected and analysed, especially during the second half of the 20th century and has shaped much of the understanding of contemporary urban theory.

2.2.2. The Social Logic of Space, and Space Syntax modelling The Social Logic of Space is a book written by Bill Hillier and Julienne Hanson published in 1984. The purpose of the book was to create a theoretical and methodological foundation for the understanding of social action in relationship to spatial structures. The core of Hillier’s and Hanson’s theory is that the study of local features of urban design and how it can be modified could give away empirical evidence about global social patterns (Hillier & Hanson, 1984 in Oliveira, 2016: p. 101). This approach suggests, just like Conzen’s study of Alnwick, that urban design and its physical layout matters for how people organize their lives. In order to study this phenomenon, Hillier and Hanson rely on the outline of street systems as a vital variable for understanding the socio-spatial compositions of cities, which has been developed into a defined theory and method for studying urban morphology known as space syntax. This approach is based on the notion that the relationship between the social and the spatial is a question of

11 human mobility and movement, which also promotes a distinct definition of how the term ‘space’ is operationalized. In the case of Hillier and Hanson, the space of cities is not a question of definitive surfaces and what they include, making them static objects – cities are rather systems of many spaces networked by human movement, thus creating a syntax and explains why economical and cultural activities manifests themselves differently in cities (ibid: p. 121). It is apparent why the study of street systems become a central component in the space syntax tradition, which is usually performed with cartographical tools such as Geographical Information Systems (GIS) where street patterns are analysed with axial-maps over cities and squares (see Figure. 1) and has become a prominent research field in urban planning, especially in regard to sustainable development.

Figure 1: example of an integration analysis with space syntax using an axial map of Stockholm in Marcus, 2008

In a Swedish context, the space syntax tradition has become popularized by researchers at The Royal Institute of Technology (Kungliga Tekniska Högskolan, KTH) and the consultancy firm Spacescape who emphasises that the concept can be understood as a measurement of urbanity, and underpin that it is a contributing factor for sustainable urban development. The founder of Spacescape, Professor Lars Marcus has been working with developing a morphological theory within the tradition of space syntax-modelling, which he refers to as spatial capital.

According to Marcus, spatial capital is an extension of the fundamental values represented by land and how it interconnects with aspects of the fixed capital, such as infrastructure, buildings and roads, as well as the spatial forms resulting from urban design. The values represented by land concerns both use-values for everyday-life and the market values (Marcus, 2008; and 2017: p. 1). Further on, Marcus presents his theory as a synthesis between social, economic, and ecological capitals, drawing inspirations from a variety of disciplines and producing a synthesized term consisting of the following components:

I. Durkheimian micro-sociology. According to Marcus, the understanding of social behaviour and its relationship to human interaction could essentially be understood by sociologist Emile Durkheim’s formulation of rituals, which are the expressions and actions of individuals that tie people together (ibid: p. 3f). The microscopic perspective

12 was also later presented by Erving Goffman’s and Randall Collins’ respective theories on interaction rituals, and ritual chains – the former emphasises that the meaning of rituals does not have be reduced to religious rites (as the original Durkheimian outline suggested) but more about banal manners and symbolisms of everyday life. The latter emphasises the role of situation contexts for rituals to be produced (ibid: p. 4). Marcus connects these perspectives to the idea of the urban capacity of generating social capital5 and bridging socially heterogeneous groups together in space. II. The new economic geography. Marcus recognizes that urban environments play a crucial role in the process of economic growth and refers to the teaching of the new economic geography by Paul Krugman that is based on two general theoretical strands; urban economics, as understood by von Thünen as land-use distributions around cities and their capacity for economic performances, and regional sciences like Christaller’s analysis of size and locations of cities and an area. Marcus concludes that spatial distributions of economic activities are related to the distribution of space, and more specifically land, which is a necessary factor for price development and is often reflected in the market values and rental costs (ibid: p. 5f). III. Ecosystem services. The last category of ecosystem services is explained by Marcus as a relatively new and contemporary process in urbanization directed by the increasing awareness of adapting cities to climate change and integrate them with social and economic developments and how to perceive and sustain them (ibid: p .6). Development strategies for urban gardening, water supplies, public transport and the generation of “green energy” et cetera, can be enabled in the urban landscape with the help spatial configurations in the urban design. However, the understanding of ecosystems differs somewhat from social and economic systems in the sense that it includes aspects such as biodiversity and the concern for life-forms other than just human and that the ecosystem services include strategies for preventing environmental damage as well (ibid: p.7).

3.0. Theoretical framework: Spatial capital and human capital and their roles in the work for urban sustainability

This chapter introduces how the theoretical concept spatial capital has been elaborated in this thesis, which differs from the urban morphological tradition in some fundamental aspects. The chapter introduces spatial capital by synthesizing elements of Pierre Bourdieu’s sociology and Torsten Hägerstrand’s time-geography. Further on, the chapter aims to connect the term with the equally ambiguous political strategy for achieving ‘social sustainable’ urban environments, which in this thesis emphasises the distribution of human capital and income in the Stockholm county. By linking the concept of sustainable development to contemporary ideas in social science about the modern metropolis as a post-industrial, or knowledge-based economy, education and specialisation becomes a more vital part of the urban identity in contrast to the industrial economy of the previous century.

3.1. Part I: The theoretical issue of spatial capital, the meaning and the measure in the context of urban and regional planning

5 According to political scientist Robert Putnam, social capital is “the collective value of all social networks and the inclinations that arise from these networks to do things for each other”, which can take the form of bonding between homogenous groups, or bridging between heterogeneous groups (Putnam, 1993 in Marcus, 2017: p. 168).

13 3.1.1. Spatial capital as a value capture From the previous chapter, it is evident that the concept of spatial capital as it appears in contemporary planning theory is seemingly a rather ambiguous one. It draws vastly upon geographical, sociological, and economical concepts and can at large be described as ‘locational advantages’ for social, economic and environmental processes being enabled through architectural design and infrastructure. Within the scope of Marcus’ perspective, the core of the argument is that spatial capital can be understood as an analytical framework for how to examine if causal mechanisms for the outcomes of social capital (in the matter integration), economic performance in relationship to space and land distributions and finally ecosystem services in regards to environmental issues are related to urban design. The outline of Marcus’ definition emphasises that spatial capital is a tool for capturing the value of land and the urban form, and that this value is equally a question of use- as much as exchange value:

“[…] exchange-value of spatial capital, suggesting how the value of urban form literally can be translated into economical capital. But just as important is the use-value of spatial capital, that is, the value urban form represents in a multitude of ways for every-day urban life, socially, culturally and environmentally. And even though not all needs request high spatial capital, in general, that is exactly what cities have been answering up to; the generic need for people and societies to access differences as a means for social, cultural and economical development” (Marcus, 2010: p. 10).

3.1.2. Spatial capital as a time-geographical concept Even if the formulation above suggests a rather broad analytical perspective, Marcus’ usage of spatial capital as a term is theorized from to the urban morphological tradition, and directly related to space syntax as a method for examining its implications. The term has however been used by other researchers within or related to urban planning, that does not rely on the concept of urban morphology. Jacques Lévy used the term in 19946 to describe the socio-spatial discrepancy between position capital and situation capital – with these two combined – the spatial capital equals to “all resources accumulated by an actor enabling him or her to benefit, according to their strategy, from using society’s spatial dimension” (Lévy & Lussault, 2003 in Barthon & Monfroy, 2010: p. 178). The dichotomy between position and situation capital is explained below:

• Position capital refers to a specific place such as the home, workplace or school. The position capital affects the individual because the neighbourhood, the social situation, location, and the image and reputation of the place will affect what types of homes, workplaces, and schools that will exist in the individual’s presence. • Situation capital refers to the chances for spatial mobility that is enabled by geographical distances and the utility of infrastructure and transportation systems. This is not equal for everybody, it varies depending on the geographical spread of transportation alternatives.

6 There is no English translation of Lévy’s works but the conceptual articles in the theoretical development includes: L’espace legitime. Sur la dimension géographique de la fonction politique (1994) and Dictionnaire de la géographie et de l’espace des societies (2003).).

14 (Barthon & Monfroy, 2010).

It is from this understanding that the thesis takes its theoretical stance.

To exemplify the meaning of position and situation capital; a place that has a high position capital means that it features high integration between social and cultural concepts such as retail, schools, workplaces and the residential areas. Inner-city environments are a god example of an ideal type for a high position capital, and suburban environments are examples of places with a low position capital. Situation capital can, in theory, compensate for the position capital by enriching the mobility and thereby minimizing the time it takes to access the high position capital environments. Poupeau et al. (2007) explains that this elaboration of the spatial capital equals to how an individual’s collection of economic, social and cultural assets is affected by the individual’s chances to access mobility. Anders Trumberg (2011: p. 120) connects in his doctoral thesis to this theoretical formulation with the traditions of time-geography7, and defines the study of spatial capital as the study of “movement in physical space”, and uses it to analyse segregation patterns in Swedish schools as a result of the freedom of school choice that was implemented in Sweden in the 1990’s. Trumberg used a combination of GIS and interviews with municipal white-collar workers and came to the conclusion that the freedom to school choice overall increased the social segregation in mid-Sweden, but with some expectations; it did not increase severely if the pupils’ families had higher access to mobility alternatives.

Sara Forsberg (2017) from Uppsala University also operationalized the term in the fashion of Lévy. She argues that the concept’s strength is that it captures the place-specific characteristics and how it relates to (the lack of) chances for mobility using Kalix in northern Sweden as an example. She conducted interviews with pupils from upper secondary schools and came to the conclusion that the spatial capital reflects the habitus of younger populations in their perspectives on future careers in geographically marginalised parts of Sweden. Plenty of students experienced themselves as ‘trapped’ in Kalix which was largely explained by the lack of mobility alternatives that relates to provincial environments. The pupils also experienced the lack of proper “lifestyle, skills, family reputation, the necessary information, school grades or contacts to enable a move elsewhere” (Forsberg, 2017: p. 17). The core of Forsberg’s research is that the cultural and social capital of individuals gets more conceptualised with the use of spatial capital as a term that can highlight geographical disparities. This can be compared to Robert Putnam’s study of Italy which stated that northern Italian cities had a higher social capital in terms of their ability of bridging (see footnote on page 13) interests between cities and regions, where as southern Italian cities had a higher social capital in terms of bonding local interests which explained the unequal development between the northern and southern parts of the country, where the North presented higher economic prosperities and trust of authority and the South presented a higher degree of clan structures and corruption (Putnam, 1994).

3.1.3. Spatial capital as a tool for urban planning Alan Mace (2017) makes another distinction of the term, relying on field-theory famously postulated by sociologist Pierre Bourdieu that in turn relies on how economic, social and cultural capital are produced under certain circumstances, metaphorically explained as a field (Mace, 2017: p. 122). The idea of the field is the expression of the networks in which and social action is enabled, where the individual’s access to capital(s) shapes their habitus (Bourdieu,

7 Time-geography is a branch of spatial science developed by geographer Torsten Hägerstrand and focuses on human movements and spatial restrictions that hinders the movements, which is largely explained as a factor of time (Trumberg, 2011: p. 88).

15 1985). The field can manifest itself under both temporal and spatial circumstances including neighbourhoods or a city, depending on the empirical scale. Mace highlights that a spatial understanding of field theory can provide urban planners with a practical theory for analysing the physical context for how different economic, social and cultural phenomena are distributed and resulting in spatial outcomes, which he labels spatial capital (Mace, 2017: p. 125). He sets down three potentials for the elaboration of the theory into a practical tool for analysis:

I. It allows planners to bring together a series of urban events under a common conceptual framework. Making connections between events could offer a more holistic insight into the operation of power in the built environment. II. It could contribute to the development of a more reflexive practice by highlighting planners’ assumptions of what is and is not given. III. It could direct planners to the distinction between place and position that highlights the limits of deterministic approaches to the built form.

(Ibid: p. 126)

Like Marcus, Mace emphasises a conceptual framework that is a meant as a theory for planning, and not a theory on planning – for example the notion of households and how they are enabled or disabled to vie for school places, exploit connectivity within and between city cores, or claim neighbourhoods for themselves (ibid: p. 130). But in contrast to Marcus’ approach of designing a theory for spatial capital on basis of urban morphology that points to the space syntax method, Mace’s suggestion of understanding spatial capital is arguably more elastic and could be analysed with more traditional tools in the social sciences.

Barthon and Monfroy (2010) for instance, applied a multinomial logit regression analysis in order to study the relationship between spatial dimensions and parents’ school choices for their children in the community of Lille (France). Their elaboration of spatial capital is inspired by Lévy & Lussault’s definition of the concept (see page 14). The study showed that the residential neighbourhoods had significant effects on the distribution of schools and pupils - the position capital was evidently a prominent factor for why certain schools were located where they were, with public and less privileged schools being dominated near working-class neighbourhoods, and more privileged private schools being dominated near middle-class areas (Barthon & Monfory, 2010: p. 183). The situation capital, which in the study concerned the aspect of mobility, was proven to have a significant effect for pupils from middle class backgrounds living in mixed or working-class neighbourhoods. 65,9 % of the pupils attended schools outside their neighbourhoods to avoid the stigmatization the schools in their vicinity were connected to.

Further on, the researchers conclude that the vicinal mobility was the most socially discriminating factor for the distribution of pupils and schools, whilst the mobility to attend privileged schools in areas further away was more developed. The situation capital thus had an impact on enriching the spatial mobility for pupils in Lille, but did not compensate for the fact that social class and spatial elements correlated – with a majority of pupils from middle-class environments living close to more privileged schools to which they attend, and pupils from working-class and mixed environments living closer to less privileged schools in which a significant share enrolled into the more privileged areas (ibid: p. 185ff). It was also established that even though social background and spatial mobility correlated, the latter phenomena almost overshadowed the former in the aspect of schooling practices. Spatial dynamics are therefore crucial factors in the question of tackling social segregation where opportunities to mobility is regarded as a fundamental need (ibid: p. 191).

16 Frenkel & Porat (2017: p. 171) on the other hand uses spatial capital as “accumulated assets and capabilities of a region” with the intention of developing a theoretical framework for strategic planning, and how to stimulate a synergy between local strategies and regional goals. They emphasise that many studies in general sometimes seek to explain development in terms of economic growth, where indicators such as GDP, income, tax rates et cetera. seek to explain the progress of a region and often neglecting local structures’ influence on the matter. According to Frenkel & Porat, regional growth is more dependent on the local structures of the social ecosystem rather than a region’s collected economic capital (ibid: p. 173). They refer to a study by Camagni & Capello (2013) that in turn aims to study the local effects of macro economical concepts by using the term ‘territorial capital’ which is used as an index that includes regional growth as well as self-employment, employment in science and technology, and transportation infrastructure. Ideally, this approach offers a more integrated perspective on how macro and micro phenomena interrelates. Frenkel & Porat’s study of the Region Valley of Springs (in Hebrew: Mo'atza Azorit Emeq Hamaayanot) in northern Israel concluded that regional development plans must be better at emphasising local capital assets and how they are spatially distributed in order to implement them successfully, and that planning theory lacks a clear empirical definition of this friction – hence spatial capital aims to capture this discrepancy and provide with place-specific knowledge (Frenkel & Porat, 2017: p. 191f).

3.1.4. Spatial capital as a theory on gentrification, new urbanism, and the rise of the creative class Spatial capital has also been addressed as a theory on gentrification by Huang et al. (2017) who also elaborates the term in the fashion of Bourdieu’s understanding and place emphasis on spatial movement and the capacity to be mobile as the conceptual parameters. This can be measured in terms of residential locations, means of transport, proximity and accessibility. A parallel to Marcus’ approach is that the researchers apply their understanding as a value capture of land which in this case estimates a certain area’s likelihood to undergo a gentrification process, using an inner-city area in Beijing as an example (ibid). The core argument of their research is the claim that spatial capital influence and institutionalize the cultural capital of places, and that this is the motive of the undergone gentrification. A similar approach was used by Rérat & Lees (2010) who also used this operationalization in order to examine gentrification strategies in Swiss cities. The distinction that both of their respective research highlights is that empirical investigations on gentrification seldom emphasise the spatial dimension, but rather the socio-economical. Both these researchers recognize the ‘urban renaissance’ – or the back to the city phenomenon of the recent decade as a contribution to gentrification. This discourse suggests that in a post-industrial economy, social and economic forces have gradually been centralized to urban contexts as a reaction against the suburbanization of the 20th century which offered sterile and homogenous living arrangements connected to expanding infrastructure (Rérat & Lees, 2010; and Huang et al., 2017). In a knowledge-based economy, that is strongly related to the concept of sustainable development (see paragraph 3.3), more ‘classic’ urban designs that can be found in city centrals that offers a lot of diversity in architectural amenities and social functions have had a strong recognition by urban planners as optimal environments for ‘sustainable living’. There is a contemporary school of thought in planning that have conceptualized these arguments into a practice that is know as new urbanism (Marcus et al., 2013: p. 88). The practice is based on the same foundations that are prominent for urban morphology. Cozens (2008) says that the arguments once made by Jacobs and Newman still matters in the understanding of social actions and its relationship to built environments, and that building cities that are designed with the capacity of maximizing the social control makes the urban experience richer and safer. Practitioners of new urbanism interpret this as a question of the degree of openness and diversity within the built environment which ideally will make

17 places more ‘active’ or ‘lively’, thus creating a feeling of inclusion in opposed to excluding designs like gated communities that would increase segregation.

Famous strategies for implementing these ideas that Cozens mention include:

• Infilling of estates to increase the physical and human densities, thus avoiding urban sprawls. • Spreading out social and economic activities within the urban fabric, thus promoting flows of people. • Integrate housing with retail on the street level, thus creating local bonds and minimizes car use for shopping. • Architectural diversity, thus creating a more colourful and vibrant atmosphere.

The concept can be recognized in many metropolitan cities around the world. In Sweden, an urban area known as Jakriborg in the municipality of Staffanstorp (outside Malmö in southern Sweden) could be seen as an example new urbanism. The area was built in the late 1990’s but makes a strong aesthetical distinction from the traditional Swedish 20th century suburbs by resembling that of a pre-industrial urban village. Places that look like these tend to be expensive and considered quite exclusive, and in cities such as Stockholm where gentrification occurs, places near or in the urban central areas tend to be targeted by gentrifiers because their existing forms and their locations makes them attractive in a city where sustainable living becomes the norm and easier to implement in these places (Dooling, 2009; and Marcus et al., 2013). In Stockholm, near-city suburbs of these characteristics such as Midsommarkransen and Aspudden that was built in the early 1900’s and traditionally have been populated by the working-class living in rental units, have largely changed their demographics and tenures form during the 21th century (Mäklarstatstik, 2018).

The reason behind gentrification from this perspective is that plenty of urban places established decades ago features a lot of architectural and spatial qualities – features that new urbanism aims to replicate, which often results in a demographic composition that Richard Florida famously proclaimed as the creative class. Florida (2002, 2006) refers to the growing trend of aspects such as digitalisation and the internet as a paradigm shift in economics and how to understand the future of growth when it becomes more based on specialisation in complex fields that require education. Meanwhile, manual labour becomes a process located in the global South, enabled by transnational policy through institutions such as the International Monetary Foundation (IMF), the EU and the World Bank. Florida also estimates that the future of Western industries lies in the “creative” working sector, and would be dominating the world economy and fiscal policies within the coming decades. This class is constituted by market innovators, business administrators, IT-professionals, researchers, scientists, and engineers et cetera. working in post-industrial urban environments in the global North (see Florida, 2002; and Roy & Ong, 2011). The geographical component in Florida’s analysis advocates that there is a relationship between economy and space, resulting in hotspots such has Silicon Valley in San Francisco where billion-dollar industries such as Google and Apple Inc. are located and is wildly regarded as one of the world’s most economically innovating places.

3.2. Part II: Social sustainability and its relationship to spatial capital

As concluded in the segments above, spatial capital is a rather ambiguous term that ought to be interpreted with a rather pragmatic caution. Whether it concerns spatial planning, gentrification or economic geography – the idea of spatial capital promotes the conclusion that space matters for social and economic activities, but also the need for developing a theoretical understanding

18 of to what degree it affects the human settlement and their life-worlds. The interpretation of spatial capital varies depending on the theoretical foundations, but the most prominent traditions that is detected in this thesis are the urban morphological approach that corresponds with architectural theory and space syntax modelling, and a more traditional social scientific approach relying on the understanding of capital in similarity with Bourdieu and Hägerstrand. This thesis will rely on the latter tradition.

3.2.1. What is sustainable development, and social sustainability in particular? The meaning of the ‘social’ in social sustainability is complex and can even be controversial. It can be argued that the term has become a ‘buzz word’ in planning documents and become something of political slogan (see Åhman, 2013). Even if the concept of sustainable development has gone mainstream, it was actually popularized in the Brundtland Commission8 in the late 1980’s as a conceptual framework for how to promote global strategies for ending poverty and increase the human well-being within the tandem of the global ecosystem (ibid). The idea of ‘sustainability’ in this context referred to the necessity for economic growth to end global poverty and at the same time emphasising the need for ecological industries to be established in order to minimize environmental damage. Sustainability as it is presented in its current discourse separates between environmental, economical, and social sustainability strategies (ibid; and Colantonio, 2011; and Dempesy et al., 2009) which is believed by some scholars to have been popularized in the 1990’s as a business strategy to overcome challenges for enterprises in order to stay competitive in rapid-changing markets (Liljefors, 2013). Since the politicization of these strategies, the question remains whether they should be practised and understood as three separate concepts, or utilized into one synergetic design (Colantonio, 2011).

Figure. 2: popularized envisioning of how sustainable development is perceived. Source: https://www.westernplanner.org/rmlui-column/2016/4/3/rmlui-legal-corner-complete-sustainability-has-three- es-does-yours

Although most scholars might argue for the latter interpretation, Figure 2 displays a strategic framework in which environmental, economic and social goals are highly detached from one another – for instance, strategies for “Workplace Development” within the Economic sphere does not necessarily need to relate to “Homelessness Prevention” or “Air Quality Programs”

8 The Brundtland Commission was previously known as the World Commission on Environment and Development (WCED) and pioneered ”sustainable development” in their report Our Common Future in 1987 (source: https://sustainabledevelopment.un.org/milestones/wced)

19 within the Social and the Environmental spheres. It could indeed be argued that these strategies contradict each other if let say “Business Retention & Expansions” make impacts on ecologies and social welfare (Keil, 2007). This problem is emphasised by Voinov (1998) as the sustainability paradox. This argument is at large concerned with the contradiction of how Earth’s resources are finite and how Western economies are (at least in theory) lacking a recognition of these fundamental aspects of the Earth’s resources. However, the understanding of social sustainability from a similar perspective can be applied compared by scholars of critical geography9 such Cook & Swyngedouw (2012) and Swyngedouw (2007) who argues that the understanding of social sustainability is that of a ‘post-political construct’ that has been reduced to technocratic processes of bureaucratic designs and enterprises, and fails to recognise ‘social sustainability’ as a political arena for implementing social justice. It is important to clarify that these scholars are inspired by a Marxian agenda that argues that social equity is in direct opposition to the process of global capitalism, which is not an argument commonly shared by all scholars. Critical thinking does however shine light on the potential theoretical issue of whether sustainability is a question of science or ideology.

3.2.2. The link between social and human capital It seems that there is a consensus that social sustainability’s main purpose is to promote social equity and minimize social disparity. Colantonio (2011) emphasise the latter, and argues that there are correlations between societal disparity and increasing crime levels, lower life expectancies and economic instabilities as a result of human beings not getting their basic needs fulfilled. Åhman (2013) and Vallence et al (2011) expresses social cohesion and the decrease of segregation as key factors. Dempsey et al., (2009: p. 292) defines social equity as a process that concerns communities that has the collective capability to sustain itself in regards to social inclusion and cohesion, with no exclusionary social qualities that hinder people to perform economically, socially and politically. This perspective could mean that social sustainability is a question of geographical scales and possibilities of promoting communities to become more equal on a local level (ibid). Social capital (se Putnam, 1994) becomes a valid concept regarding the practice of sustainability strategies since it refers to relational aspects to strengthen the collective through bonding (i.e. the relations between individuals in a community) or bridging (i.e. the relations between two or more communities). The understanding of the value of social capital has been expressed within the theoretical model of spatial capital regardless of it being expressed as a theory of urban morphology or as extension of Bourdieu’s capital forms or Hägerstrand’s time-geography. The understanding is that spatial and social capital are related in theory. It has also been noted in Part II of this chapter that the contemporary trends in urban and regional planning (such as new urbanism) are based on an economic framework of a knowledge-based, post-industrial paradigm for achieving economic growth, where what Richard Florida (2002) describes as the creative class are becoming the ‘avant-garde’ of this ‘sustainable’ lifestyle and therefore highlights the concept of human capital.

To put in bluntly, human capital is an individual’s set of specialised abilities and competences that are requested on the contemporary labour market; this includes a spectrum of attributes such as level of education, intelligence, creativity, personality traits et cetera (Becker, 1964). The term was popularized by the Nobel prize-winner and economist Gary Becker (ibid) in the book Human capital: a theoretical and empirical analysis, with special reference to education from 1964. The argument Becker made was that employers are likely to invest in individuals that are likely to generate higher turn-overs and strengthen the business-culture through

9 Critical geography is understood here as an elaboration of critical theory and other postmodern attempts for detecting (potential) power asymmetries in and between different geographies.

20 individual personality traits and specializations. He emphasised that higher education was a crucial parameter for achieving a higher human capital (ibid). Plenty of social scientists acknowledge human capital as a fundamental theory in urban economics, especially when it comes to achieving sustainable growth (see Rauch, 1991; Florida, 2006; Weingaertner & Moberg, 2011; and Storper & Scott, 2012). A general consensus is that the urban environments such as central business districts, where urban growth stems from have certain social and cultural features, or amenities, that one needs a high degree of human capital to access (Storper & Scott, 2012: p. 148). Once again this emphasise a relationship between the enablement of the ‘sustainable’ planning paradigm and spatial conditions that are required.

There are however different theories of how to best understand the human capital’s role in the growth of cities. Richard Florida’s theory on the creative class is one example, but the economist Edward Glaeser (1992, 2004) rather emphasises how amenities of urban areas that consists of a great deal of architectural and cultural diversity, usually apparent in north-eastern American cities have benefited from a resurgence when economic growth now is more dependent on an increase of human capital and knowledge-based industries. In short, since the mid 1940’s most of the United States’ most prominent economic strengths has been located in the Sun Belt10 due to dry winters, cheap housing costs and tax revenues, sprawling suburbs to strengthen local identities, and the fact that implementation of motor-highways and airports did not make metropolitan cities dependent and railways, coastlines, large lakes or rivers (ibid). During the 20th century north-eastern cities such as New York City and Boston went through a population decline between the years 1950 to 1980 and economic stagnation which resulted in social problems, and in New York’s case bankruptcy in the mid 1970’s. According to Glaeser, this has changed due the increase of human capital on the labour market, and that young professionals of today a more attracted to north-eastern cities due to their rising amenity values and the density which allows for cultural diversity (consumer-cities, according to Glaeser) The more traditional ideals of ‘city-life’ that can be recognized in the teachings of Jane Jacobs are also more optimal for developing urban sustainability due to their compactness and inter- connectivity (ibid; and Storper & Scott, 2012: p. 149ff, p. 161).

3.3. Theoretical conclusion: The geographical link between urban sustainability in the form of human capital and income is spatial capital

The conclusion of this chapter is that the even though the meaning of the ‘social’ in social sustainability is theoretically and politically diffuse, international research highlights a need for minimizing social segregation and disparity as well as promoting integration and equity. It can be argued that achieving this is a political goal, and therefore no strict scientific terminology can be requested in its adaptation. There is however a strong notion presented in the chapter; the promotion of socially sustainable cities is understood as the strengthening of the social capital – that is the social bridging between places and bonding within places, as well as increasing the human capital – meaning individual competencies – through education to achieve economic growth. This is believed by some scholars to achieve higher equities and generate well-being (see Glaeser et al., 2004; and Weingaertner & Moberg, 2011). It is also understood that the spatial conditions matters for human settlements, economic performances and social life, which is the core principle of the use of the concept of spatial capital regardless of its application in different disciplines. In theory, a place with a high spatial capital in the form of position or situation capital should be ideal in the development of ‘sustainable’ cities.

10 The Sun Belt refers to the southern states in the USA, such as California, Arizona, Texas, Florida et cetera.

21 Both the urban morphological approach as well as the research influenced by Lévy’s conceptualization emphasises social capital as an important component of the spatial capital theory. Human capital on the other hand is seemingly a blind spot in both these frameworks, which is rather interesting considering its role in urban growth in post-industrial economies and popularisation by prominent urban theorist such as Becker, Florida and Glaeser. This opens up for empirical investigation on whether spatial capital, (understood as either position or situation capital), and human capital (understood as higher education), as well as high incomes are correlating phenomena in a supposedly ‘sustainable’ city.

4.0. Methodology: quantitative design and hypothesis testing

In order to answer the research questions and examine whether the theoretical framework can be supported by empirical evidence, the thesis is based on a case study using Stockholm county as the object of research. This chapter presents the general outline and research design, what methods were being used and how the theory was elaborated. Further on, it also includes an epistemological discussion concerning the role of the hypotheses being used and how to interpret the initial results of the study. It also includes a discussion on how to evaluate the sources and research ethics in social science.

4.1. Research design

The study is a consists of a quantitative research design, combining Geographical Information Systems (GIS) with multivariate regression analysis. The motivation for this approach is that both GIS as well as regression analysis are powerful tools of applied mathematics used for scientific experiments and testing hypotheses. GIS-methods are based on cartographical designs that allows one to study how statistical data are spatially distributed on a world map and has become a valid instrument for geostatistical modelling in recent decades (Fotheringham et al., 2000: p. 30f). Regression analysis is a widely used statistical method applied in many disciplines across the social, physical, behavioural, and life sciences. The purpose of the method is the estimation of the strength of a correlating relationship between an independent, X, and dependent, Y, variable with the help of a reference line in a Cartesian coordination system (Djurfeldt et al., 2013: p 157). This approach regenerates an empirical framework for testing hypotheses and suggesting possibilities for causal inference by drawing conclusions based on parameter estimations on randomised samples.

4.1.1. Operationalization of theory and data used The research design follows a deductive logic, meaning that the theoretical components introduced in the previous chapter are meant to be operationalized into empirical (in this case measurable) units that can generate evidence that the theoretical model provides significant truths-claims about the real world. The relationship between theory and empirical evidence can be a philosophically controversial discussion in social science. The American sociologist Robert K. Merton (1949) argued that the the nature of many prominent social scientific theories are to ‘grand’ and impossible to utilize into one conceptual idea to be scientifically tested. Thus, Merton suggests that in order to conduct social science that generates any empirical findings, theories must provide with an abstract model that can be transformed into concrete hypotheses that stands up for tests from which the researcher can draw conclusions about a theory’s validity on a middle-range level (ibid; and Boudon, 1991; and Bryman, 2011: p. 26).

The rationale of this approach suggests that finding evidence for grand social theories to be universally accurate are deemed impossible from an empirical perspective, but rather the

22 aspects of a theory can be proven. Therefore, this study has been conducted using the middle- range concept, with the theoretical framework of the relationship between spatial and human capital functioning as an ideal type operationalized into variables that can either confirm of reject the hypotheses based on the findings. This means that the quantification of the theories consists of proxy-measurements that examines the theories’ claims and determines its validity. Since the purpose of this thesis is to test if spatial capital (understood here as a dichotomy between position and situation capital) correlates with the spatial distribution of the human capital (understood as tertiary education), as well as the share of individuals with disposable incomes – the variables and their sources are depicted in the matrix below:

Variable Theoretical disposition Unit of measurement Source

X1: Distance to Spatial capital (position) Geographical proximity (Own calculation) Stockholm’s Central in metres Business Districts (CBD)

X2: Distance to regional Spatial capital (position) Geographical proximity RUFS, 2010 urban centres in metres

X3: Distance to nearest Spatial capital (position) Nearest distance in Open Street Map, 2016 amenity (including metres cinemas, theatres and museums)

X4: Distance to nearest Spatial capital (situation) Nearest distance in Open Street Map, 2016 railway (including train, metres metro and tram stops)

X5: Distance to nearest Spatial capital (situation) Nearest distance in Open Street Map, 2016 bus stop metres

Y1: Tertiary education Human capital Share of persons age 25- ResSegr /SCB, 2012 64 years old who have completed tertiary education divided on two grid sizes, 1 =equals urban area, 2 = rural area Y2: Taxable earned Income Share of persons age 25- ResSegr /SCB, 2012 income 64 years old who have a level of taxable earned income in the highest decile divided on two grid sizes, 1 =equals urban area, 2 = rural area

Variable X1 was calculated by creating a polygon in the GIS software for Gustav Adolfs torg in central Stockholm. Variable X2 was provided by www.rufs.se, which is produced by The department of Growth and Regional Development at the Stockholm County Council. The regional urban centres include: Barkabry-Jakobsberg, Skärholmen-Kungens Kurva, Haninge C/Handen, Täby-Arninge, Södertälje, Flemingsberg, Kista-Sollentuna and Arlanda. The variables X3 – 5 were provided by www.openstreetmap.org which is an open source of global cartographical information and provides with data on streets, roads, amenities, railway stations et cetera. It is noted that Open Street Map is updated by its users and not always comprehensive in its provided data (see page 29). Dependent variables Y1 and Y2 were provided through the research project Residential Segregation in Europe (ResSegr) who used the same variables to

23 analyse urban segregation in several European countries. The variables measures the share of people with tertiary education and the ones that have a taxable earned income in highest decile amongst the 400 nearest individuals between the ages 25 to 64, using k-nearest neighbour (k- NN) in the software EquiPop11 (Nielsen et al., 2017: p. 59). Their data is based on information from Statistics Sweden (SCB). It is important to mention that the variable Y2 for taxable earned income is based statistical data on whole of Sweden, but only presents data on Stockholm county in this study. It is also important to emphasise that the variable only measures the distribution of individuals with high incomes (that is in the highest decile) The coordinates in this data are aggregated to 250 m2 in urban areas12 and 1000 m2 outside urban areas, positioned in such away that 16 urban grids can fit in one rural grid (ibid).

4.1.2. Research object and cartographical data The study was conducted on Stockholm county which covers a total land area of 6 516 km2 and as of March 2018 has a population of 2 308 143 people in total and a population density of 354 inhabitants/km2, making it the most populous county in Sweden (SCB, 2018). It contains 26 municipalities including the city of Stockholm which is the Capital city of Sweden. A significant share of the county’s geography consists of islands, making it an archipelago. For the GIS analysis; the reference system for the coordinates was RT90 2.5 gon V (EPSG: 3021). The Backgrounds in all GIS maps were provided by ©Google Maps. The maps on the next page illustrate the all the variables used in thesis. For fuller explanation of the cartographical content see APPENDIX.

Map 1: Stockholm count.: source SCB. Background: © Lantmäteriet Geodataverksamheten: http://geodata.scb.se/reginawebmap/main/webapp/?typ=lan&l=01&a=0000)

11 EquiPop is a software developed by the geographer John Östh and is used for statistical research using k-NN in large datasets. Source: http://equipop.kultgeog.uu.se/Tutorial/Introducing%20EquiPop.pdf 12 Tätort in Swedish, which is defined as a geographical area consisting of a minimum of 200 buildings with a maximum of 200 metres apart, and no more than 50 holiday cabins.

24

Map 2: the regional urban centres of Stockholm county, Scale: 1: 250 000 Map 3: all bus stops in and to the Stockholm county. Scale: 1: 500 000

Map 5: all railway stops in the Stockholm county: Scale 1: 339 000 Map 6: all amenities in the Stockholm county: Scale 1: 330 000

Map 7 & 8: geographical proximities between all dependent variables and tertiary education in black, and independent variable taxable earned income in blue. Scale 1: 160 000

4.2. Execution

In order to conduct the statistical analysis, the computer software QuantumGIS (QGIS) and SPSS Statistics were used. QGIS is a free-of-charge GIS-software that offers plenty of cartographical tools for vector and raster projections, where the user works with the variables

25 in forms of visual layers. The software ArcMap was also used in order to produce Map 9-12. SPSS is a software for statistical analysis, owned by IBM that offers a variety of mathematical methods for statistical analyses. This study has been conducted using the following methods and tools in each software:

4.2.1. Geographical proximity analysis This was done with the help of the Distance matrix command, which is an algorithmic function that creates a table containing a matrix of distances between all the points in a points-layer. This operation was used to calculate the linear distance (in metres) from one object to another. In this thesis, this tool was used in order to elaborate the spatial capital theory to measurements of proximities and in turn calculate the linear distance from one spatial capital unit (the target data) to the distribution of the share of tertiary education as well as the taxable earned income (input data). For example, in the case of using Railways as target data, this command calculates the distance from the share of individuals with tertiary education from every unique tertiary education ID to the 1 nearest Railway ID. Also, each linear distance from a target to an input ID was modified using the natural logarithm, ex (Ln) in order to decrease the distance measurements for the highest values, or outliers, meaning that they will not drastically distinguish from the arithmetical mean-value of the population. It is preferable when performing a regression analysis that the data is normally distributed around the mean-values, thus avoiding statistical outliers that affect the parameter estimations. The reason for the proximity analysis is to provide with descriptive data on distance-values and visualised using a map. It also provides with real numerical data that can operate on a continues interval scale, which is the scale that a linear regression analysis presumes and requires in order to be efficient.

4.2.2. Stepwise multiple linear regression analysis After the GIS analysis was conducted, SPSS was used in order to estimate whether the (Ln)distances correlates with the distribution of the human capital and the income and to examine if the correlation fits a linear slope by using multivariate regression analysis. The purpose of this method is se if correlating phenomena can be explained by a line which estimates the strength of the correlation and its direction and determines whether it is statistically significant or not. In a linear regression model, the goal is to find a correlation between X and Y that resembles the mathematical precision of a line that is explained with the formula: y = α + βx

Where a is the intercept (the value of Y when X = 0) and β is the slope (Djurfeldt et al., 2013: p. 355). In the actual analysis, it is required that the regression is being performed by calculating the standard errors (εi) by using Ordinary Least Squares (OLS) in order to minimize the residuals, that is the distance between the regression line and the actual observed units, thus making the formula for regression analysis:

yi = α + βxi + εi

A multiple, or multivariate, linear regression analysis is a tool that tests two or more independent variables’ effects on one dependent variable and determines which independent variable that has the strongest effect. The mathematical equation that explains this is:

yi = α + β0 + β1x1 + β2x2 + […] + εi

26 This study used a stepwise command which is a feature that excludes variables that do not display any statistical significance. In order to interpret a multiple regression model, certain statistical elements require a more detailed understanding of the terminology being used in the analysis. In the next section, nine central concepts are explained that are of significant importance when conducting a multivariate regression analysis.

4.2.3. Key statistical concepts and meanings Coefficients Are the parameter estimations. The different coefficients determine how effective the independent variables are in predicting the a dependent variable. R2-value R2 (R-square) is a coefficient that determines the strength of the regression by providing a 0-1 ratio that examines the variance of the change between the dependent variable that can be explained by the independent variable. A higher R2- value (1 is maximum) predicts a stronger regression model. 95 % confidence An interval providing with a 95 % probability range that the parameter estimations interval represents the ‘true’ estimations outside the sampled data. Statistical significance indicates that these criteria are satisfied. Beta-value This is a coefficient that determines the direction of the slope of the fitted line and whether the correlation is positive or negative, thus estimating the expected outcome of the dependent variable when a unit on the independent variable increase or decrease. P-value A crucial coefficient that estimates whether the results are statistically significant or not. In regard to the confidence interval being 95 %. The P-value must be = 0.05 or lower: P = 0,05 – 0,01= a significance of * P = <0,01 – 0,001= a significance of ** P = <0,001 = a significance of *** Residuals Also known as standard errors. In a OLS-regression model, the residuals are the deviations between the predicted values of the dependent variable (determined by the slope of the line) and the observed values gathered from the empirical material. Higher distances means weaker predictions. Heteroscedasticity A common phenomenon that can occur in a regression model when the data is not normally distributed around the arithmetic mean-value. This would mean that the variance of the standard errors is not constant, making it difficult to estimate a linear relationship between the variables. This is controlled for by standardization of the predicting variables and the residuals by using the ZRESID function for the dependent variable and ZPRED function for the independent variable in SPSS. Multicollinearity A situation when two or more independent variables are strongly correlating with each other, thus creating a redundant analysis that makes it impossible to separate the effects of the independent variables on the dependent variable. This can be diagnosed by SPSS by checking for tolerance-values and the Variance of Inflation Factor (VIF). Auto-correlation This concept means that there is a lack of independence in the data that the variables are measuring. This is common when the data is not random but deduced from the same sources and might overlap in time and/or space. This is tested for by using the Durbin-Watson command in SPSS. (Chan, 2004; Sundell, 2009; and Djurstedt et al., 2013).

4.3. Hypothesis testing and methodological strengths and limitations

In order to answer the research questions, a set of hypotheses that are all elaborated from the rationale of the theoretical model were tested. A hypothesis works as statements of expected outcomes, and the testing of the hypothesis is necessary in order to examine whether the statement is supported by empirical evidence (Bryman, 2011). According to Arthur L. Stinchcombe, the contemporary trends in quantitative social science is to “reject the null-

27 hypothesis” which is here understood as the opposite of the hypothesis’ claim (Stinchcombe, 2005: p. 239). This emphasises the usage of regression analysis as a valid instrument for social research because the logical design of its methodology is the use of probability theory in order to estimate the statistical likelihood of the results being or not being coincidental. Further on, the larger role of the hypothesis in the relation to the theoretical framework from which its has been elaborated is a complex matter. Stinchcombe highlights that if a hypothesis cannot be empirically verified, does not automatically disqualify the theory from which it originates, but rather emphasises limitations relying on this type of research approach (ibid: p. 240f). This has already been emphasised by Merton’s argument that epistemological truth-claims in social science should be understood on a middle-range level. Another limitation of using statistical methods of this sort is that regression models estimates correlations in relation to slopes, and correlation does not imply causation (see Aldrich, 1995), meaning that it is philosophically controversial to claim that Y is an effect that is believed to have been caused by X, when correlations only estimates a relationship that logically can go both ways and can also be auto- correlated. A regression model however, should in theory minimize this dilemma because it fits the correlation’s direction to the mathematical equation of a line, thus the analysis determines the strength of this direction.

There is also a methodological controversy on whether a hypothesis testing should aim for verification or falsification. If one aims to reject the null-hypothesis the former is preferred, but according to Karl Popper’s seminal work The Logic of Scientific Discovery (1959), falsification is of higher virtue for hypothesis testing. The reason is that Popper consider it impossible to make any epistemological truth-claims based an observable evidence, but only claims about false statements. The principle of this approach is related to the problem of induction13 and the argument that observed data does not postulate information on causality – just because something has occurred multiple times before does not necessarily mean that it will again in the future. If accepting this, it is logically impossible to verify that any hypothetical outcome is true, but it is possible to verify if an outcome is false. Falsifiability should in theory provide with a stronger scientific validity than verification, which according to Popper would mean that the rejection of a hypothesis made by the researcher would force him or her into a position where (s)he can be proven wrong, and new knowledge can be detected and provide with critical insights on a theory, instead of the narrower empiricist tradition of verifying14 their hypotheses which only results in continues confirmations (ibid).

In this thesis, the research design takes influence from Merton’s criterion for deductive research modelling that estimates hypotheses on a midrange level, as well as Popper’s criterion for falsification. It is based on an elaboration of five hypotheses (H1-5) deduced from the theoretical implications, each with a contradicting null-hypothesis (H0). If a hypothesis is successfully falsified it will be understood as a confirmation of the null-hypothesis. As emphasised in the next section, the expecting results will be negative correlations between the spatial capital and the human capital and as well as the income.

4.3.1. Hypotheses The hypotheses tested in the thesis were the following:

13 This is a philosophical understanding that are related to the teachings of the 18th century philosopher David Hume which suggests that a countless number of observations being made cannot justify them as causal (source: https://plato.stanford.edu/entries/induction-problem/) 14 In this thesis this refers to the teachings of the Vienna circle and the epistemological framework of logical positivism (source: https://plato.stanford.edu/entries/vienna-circle/)

28 H1: “Shorter distances to the city business district (CBD) and regional urban centres will both have the strongest effects on tertiary education and taxable earned income” • H0: “Distances to the city business district and regional urban centres will not have the strongest effect on the share of tertiary education nor taxable earned income” H2 “Shorter distances to the city business district will predict tertiary education and taxable earned income stronger then any other factor” • H0: “Distances to the city business district will not be the strongest variable” H3: “Shorter distances to railways, bus stops and amenities will correlate with a higher share of tertiary education and earned taxable income” • H0: “Neither distance will predict tertiary education and taxable earned income” H4: “Distances to railways will predict tertiary education and taxable earned income better than distances to bus stops” • H0: “Distance to railways and bus stops will not predict tertiary education and taxable earned income” H5: “Distances to amenities will have the weakest effect on tertiary education and taxable earned income” • H0: “Distance to amenities will not have a weaker effect than the other independent variables”

4.3.2. Ethical considerations and evaluation of sources Ethical considerations in research for the humanities and the social sciences are consisting of four main principles set down by the CODEX programme by the Swedish Research council (Vetenskapsrådet, VR); criterion for information, criterion for consent, criterion for anonymity, and criterion for usage (VR, revised in 2018). These principles are more oriented toward research methods that require human interaction. Since this thesis’ research strategy focuses on aggregated data, these principles are of lesser concern. The empirical data consists of public information deducted from official sources, with the exception for the underlying material for tertiary education and taxable earned income which was gathered from a previous study done by Nielsen et al (2017) with the approval from one of the authors. The author of this thesis hereby guarantees that no plagiarism has taken place during the production of the thesis.

When conducting social research one should also always be cautious with the material being analysed. In this thesis the GIS-based data for the variables X3-5 that are operationalized as spatial capital units were provided through Open Street Map which is an open database that allows anyone to access and edit global geographical information on roads, amenities, railways et cetera. The validity of such a source can be be questioned – however, considering the fact that the information that the service provides is not socially controversial or classified, it is presumed that the information is correct. Open Street Map is also a partner of several universities and organizations which indicates that it is a serious source (source: https://www.openstreetmap.org/about). The rest of the content of this thesis has been deduced from several official sources, including the Swedish agencies: Agency for housing (Boverket), Statistics Sweden (Statistiska centralbyrån, SCB), and the Management for Economic Growth and Regional Development at Stockholm county (Tillväxt- och regionplaneförvaltningen at Stockholm läns landsting). The theory literature presented in the chapters 2 and 3 consists of sources from academic anthologies and peer reviewed articles published in academic journals. The literature was mostly provided through Stockholm University’s library service, the author’s self owned literature, and through Google Scholar.

29 5.0. Results

This chapter presents the empirical findings of the study by introducing the results from the proximity analysis that was performed by using the GIS-software QGIS and from the regression analysis performed in SPSS. The chapter starts with the descriptive statistics, histograms and scatterplots of the regressed variables. It then presents a summary of coefficients, collinearity diagnostics and auto-correlation, maps over the residuals, the change of the coefficients when excluding geographical proximity to the central business district, hypothesis testing and finishes with a summary of the empirical findings.

5.1. Proximity analysis and descriptive statistics

The proximity analysis was conducted with the help of the distance matrix function in QGIS. The data for both dependent variables concerned the year 2012. The statistical elements are presented in decimals as well as in quintiles consisting of 5 classes (see Map 5 & 6). This means that the GIS is dividing the shares of both variables into five categories of equal statistical spread. The cartographical data displays a strong geographical pattern, where both high values for tertiary education as well as taxable earned income densifies in and near the CBD, and decreases on the outskirts where both education-levels and incomes are visibly lower. The values are also higher around the regional urban centres and where distances to nearest railway, bus stop and amenity are seemingly shorter. In order to test for statistical significance with a linear regression model, it is expected that all variables are featuring real number-data that can be measured on a quota/interval scale (Djurstedt et al., 2013: p. 355ff). It is therefore important that the observed units are normally distributed around the arithmetic mean value (µ). Table I and Figure 3: the histograms below are presenting the descriptive statistics for each dependent variable and their distribution.

Table I: descriptive statistics of variables Mean (µ) Std. Deviation (σ) Observations (N) Y1: Tertiary 0,4277077 0,155678285 15 802 education 72

Y2: Taxable 0,1767920 0,112785657 15 802 earned income 2

X1: Distance to 9,8580 0,78686 15 802 CBD X2: Distance to 7,3078 4,94924 15 802 regional urban centres X3: Distance to 8,3662 1,06040 15 802 nearest amenity X4: Distance to 7,8642 1,43823 15 802 nearest railway X5: Distance to 5,9810 1,21561 15 802 nearest bus stops

30

Figure 3: histograms displaying the distrubution around the artimethic mean-value

The total amount of observed units (N) were 15 802 for all observations and were all farily even in its distribution according to the applied normal distribuition curve, which is most likely due to the use of the natural logarithm and the standarization on the variables (see next paragraph) that hypothetically would result in statisical outliers that would have distorted the distribution. It is clear that the observations for tertioary education are closer to the mean value than for taxable earned income, which displays a slight tilt towards one standard deviation from the mean-value . However, all of the observations are within the 95 % probability reach.

5.2. Standardized residuals

The residuals can be explained as the distances between the observed data and the linear model that determines the estimated strength between the variables. Ideally, one wants the observations ‘glued’ to the reference line to minimize the standard error frequencies. When the parameters are estimated using ordinary least squares (OLS), the minimization of the sum of squared distances parallel to the y-axis (between each observed unit and the corresponding point on the regression line) is of preference. The residuals are of high importance and analysis that comes with certain presumptions – this includes that the standard error for each observation is independent and that the errors do not correlate with each other. There is also as the presumption of homoscedasticity. If a regression model demonstrates heteroscedasticity it would mean that the residuals are heterogenic in its distribution and could possibly affect the strength of the estimation, making it difficult to predict the relationship. This was controlled for by choosing to standardized the residuals, which means that SPSS automatically sets the arithmetic mean- value to 0 before computing the regression weights (see histograms above). Standardization enables a more solid platform for comparing weights, thus making the coefficients less likely to predict heteroscedasticity (ibid: p. 367). In SPSS, this is controlled for by producing scatterplots using the ZRESID command for the dependent variables which checks for linearity, homoscedasticity and statistical outliers by standardizing the residuals, as well as the ZPRED command for the independent variables which standardizes the predicted values (Sundell, 2009; and Djurfeldt et al., 2013).

31

Figure 4: Expected cumulative probability for dependent variables tertiary education and taxable earned income

Figure 4 examines the cumulative probability plots (or p-plots) of the residuals and displays that the variables used in the regression are following a general pattern of homoscedasticity. Especially for tertiary education, but seemingly less so for taxable earned income. This can easily be recognized by the ‘tightness’ of the plots and how well adjusted they are to the reference line. Figure 5 (on the next page) is proving with a visualisation over the OLS modelling using standardization for each independent variable’s effect on the dependent variables by producing scatterplot-diagrams. As understood from Figure 4, the linear regression model predicts dependent variable tertiary education somewhat stronger than taxable earned income (it is called disposable income in the scatterplots). The overall relationship does imply very little deviation from the predicted values. Figure 5 (next page) displays the relationship between each independent variable and the dependent variables by adding a reference line.

32

Figure 5: scatterplots for regressed independent variables affect on dependent variable tertiary education to the left, and earned taxable income to the right

33 5.3. Summary of coefficients

All independent variables were proven to have statistically significant effects on tertiary education as well as taxable earned income. The confidence interval being 95 % requires a P- value of 0,05 or less in order to be significant. The tables III and IV are presenting the effects of each independent explanatory variable on the dependent variables. The numbers (1) – (5) are representing the numbers on each regression model. Since this analysis was conducted with the use of the stepwise command, the analysis only includes the models that are of significance, which in this case includes all five. The variable X1: Distance to CBD was the variable that had the strongest effect in both analyses (see R2-values in Table II and III). The first numbers on the rows in the tables presents the unstandardized beta-values that determines the direction of the correlation and the level of significance. The subtraction symbol indicates if it is a negative correlation and the *-symbol the level of significance which for all five variables were P = ≤ 0,000. The standard deviations for the independent variables as well as the intercept which in SPSS is called “Constant” are presented in the parentheses. The R2-value presents how much 2 of the variance of Yi that can be explained by a variance Xi for each model, and the R -change displays how much the value increases or decreases between models.

Table II: coefficients for dependent variable tertiary education (1) (2) (3) (4) (5) Distance to -0,140*** -0,153*** -0,159*** -0,154*** -0,156*** CBD (0,001) (0,000) (0,002) (0,002) (0,002)

Distance to 0,004*** 0,003*** 0,003*** 0,003*** regional (0,000) (0,000) (0,000) (0,000) urban centres

Distance to 0,009*** 0,010*** 0,010*** nearest (0,001) (0,001) (0,000) amenity

Distance to -0,005*** -0,006*** nearest (0,001) (0,001 railway

Distance to 0,005*** nearest bus (0,001) stop

Intercept 1,802*** 1,912*** 1,898*** 1,878*** 1,886*** (0,011) (0,013) (0,013) (0,013) (0,013) R2 (adjusted) 0,497 0,506 0,508 0,509 0,510

R2-change 0,497 0,009 0,002 0,001 0,001

N 15 802 15 802 15 802 15 802 15 802 P = ≤ 0,05*; P = ≤ 0,01**; P = ≤ 0,001***

34 Table III: coefficients for dependent variable taxable earned income (1) (2) (3) (4) (5) Distance to -0,063*** -0,089*** -0,101*** -0,106*** -0,109*** CBD (0,001) (0,001) (0,001) (0,001) (0,002)

Distance to 0,007*** 0,006*** 0,006*** 0,006*** regional (0,000) (0,000) (0,000) (0,000) urban centres

Distance to 0,019*** 0,018*** 0,007*** nearest (0,001) (0,001) (0,001) amenity

Distance to 0,003*** nearest (0,001) railway

Distance to 0,018*** 0,006*** nearest bus (0,001) (0,001) stop

Intercept 0,801*** 1,002*** 0,973*** 0,989*** 1,002*** (0,010) (0,011) (0,011) (0,011) (0,012) R2 (adjusted) 0,195 0,254 0,271 0,275 0,276 R2-change 0,195 0,059 0,017 0,004 0,001 N 15 802 15 802 15 802 15 802 15 802 P = ≤ 0,05*; P = ≤ 0,01**; P = ≤ 0,001***

As evident in Table II, distance to the CBD and distance to nearest railway display negative beta-coefficients for the regression slope between tertiary education. The table displays positive beta-values for distances to regional urban centres, nearest amenity, and bus stop. The correlation matrix that comes with the regression analysis does however inform of negative correlations for each of the five variables (see APPENDIX). It is necessary to highlight that correlation- and beta-coefficients are not the same things and do not display the same information; the former is only a measurement of the statistical relationship between the variables whilst the beta-values measures the strength of the slope in order to estimate the correlations continues direction (Djurfeldt et al., 2013: p 262). The unstandardized beta-values remain somewhat stable between all five models, with the exception of distance to CBD which changes somewhat between models.

The adjusted R2-values for each model stretches between 0,497 – 0,510 which means an overall increase with 0,013 on R2-change. Each model thereby determines that the models explain the relationship by roughly 50 % for each model. The reason for choosing to present the adjusted values is that in a multiple linear regression analysis, the R2-values for a model tend to increase when adding variables and thereby enables the researcher to manipulate the outcome by adding variables that do not fit the assumed linear model, thus pressing up the determination coefficient for a model resulting in a possibility of biased results (ibid: p. 161). The adjusted R2 compensate for this by only adding useful values to the model summary and disqualify those that are decreasing (ibid). In this model the adjusted R2-values are only increasing.

35 Table III displays similar features for dependent variable taxable earned income, with all the regressed variables being significant at a ***-level. The distance to the CBD is the only variable that displays negative beta-values. One interesting detail is that distance to nearest railway was proven to have zero effect in model (4) and thereby disqualified by the stepwise command. Distance to nearest railway did not have a significant effect on the taxable earned income according to the model, but has in model (5). The adjusted R2-values are weaker for this dependent variable with each model predicting the taxable earned income between circa 20 – 28 % each.

5.4. Collinearity diagnostics and auto-correlation

A common problem with multiple linear regression analysis is the possibility of multicollinearity (see Chapter 4). There is no absolute consensus on how strong the correlation between the independent variables have to be and still be regarded as acceptable, but according to Sundell (2009), correlations stronger 0,080 are usually problematic. In this study there are no correlations of that strength between any of the independent variables (see Correlation matrix in APPENDIX). SPSS allows the testing for multicollinearity through the option “Collinearity diagnostics” which provides with two collinearity measurements; tolerance and Variance Inflation Factor (VIF), with the latter being equal to 1 divided by the value of the former. According to Sundell, there is no definitive number that determines when a regression model is affected by multicollinerarity, but a common approach is a range from 1 – 5 for the VIF-value (ibid). Djurfeldt et al., (2013: p. 366) claims 2,5. This study has been conducted with the VIF criterion for multicollinearity being > 3, making the tolerance value ≈ 0,33. Ideally, one would want the tolerance being as high as possible where 1 is maximum, and the VIF-value being as low as possible where 1 is minimum (ibid). The Durbin-Watson test checks for auto- correlation which means a that there is a correlation between the values of the same variables appearing at the same spatial or timely dimension. The test operates on a range from 0 – 4, where values around 0 indicates a strong positive correlation, values around 4 indicates a strong negative correlation (Chan, 2004). The model summaries for both tertiary education and taxable earned income were < 1 (see APPENDIX) which indicates a strong positive correlation and violates the independence of the variables. This means that the residuals in the models have an over predicament which is most likely due to the fact that a significant majority of the observations are clustered in in or near the CBD.

5.5. Exclusion of X1: Distance to the CBD

There were no VIF-values as high as 3 or higher in this study; the highest being the score for distance to the CBD in model (5) for both regression analyses. Since the study was conducted with the hypothesis that the geographical proximity to the CBD would explain the share of people with tertiary education and taxable income better than any other variable, there is a chance that the other independent variables’ correlations with the dependent variable are spurious, meaning that their explanatory effects would be weaker due to the presence of a stronger predictor. In the correlation matrix’ for both tertiary education and taxable earned income, the variable X1 is proven to have the strongest correlation to both dependent variables (see APPENDIX). This is a problem when testing the other independent variables’ effects on the dependent variables. therefore, the multivariate regression was performed again, this time with the variable X1 excluded.

36 Table IV: coefficients for dependent variable tertiary education excluding distance to CBD (1) (2) (3) (4) Distance to -0,001*** regional (0,000) urban centres Distance to -0,025*** -0,022*** -0,020*** nearest (0,001) (0,001) (0,001) amenity

Distance to -0,057*** -0,047*** -0,044*** -0,043*** nearest (0,001) (0,001) (0,001) (0,001) railway Distance to -0,009*** -0,009*** nearest bus (0,001) (0,001) stop

Intercept 0,874*** 1,005*** 1,012*** 0,996*** (0,006) (0,008) (0,008) (0,010)

R2 (adjusted) 0,276 0,296 0,299 0,300

R2-change 0,276 0,021 0,003 0,001

N 15 802 15 802 15 802 15 802 P = ≤ 0,05*; P = ≤ 0,01**; P = ≤ 0,001***

Table V: coefficients for dependent variable taxable earned income excluding distance to CBD (1) (2) (3) (4) Distance to 0,002*** 0,002*** 0,003*** regional (0,000) (0,000) (0,000) urban centres Distance to -0,002* nearest (0,000) amenity

Distance to -0,021*** -0,025*** -0,023*** -0,023*** nearest (0,001) (0,001) (0,001) (0,001) railway Distance to -0,004*** nearest bus (0,001) stop

Intercept 0,343*** 0,357*** 0,369*** 0,382*** (0,005) (0,005) (0,005) (0,008)

R2 (adjusted) 0,072 0,079 0,080 0,081

R2-change 0,072 0,007 0,001 0,000

N 15 802 15 802 15 802 15 802 P = ≤ 0,05*; P = ≤ 0,01**; P = ≤ 0,001***

37 The relationships between all independent variables and each dependent variable are displaying a negative trend when excluding the variable geographical proximity to the CBD, except for distance to regional urban centres in relationship to taxable earned income. In the case of both tertiary education and taxable earned income, distance to nearest railway is significant at P = ≤ 0,000*** for all four models and remains somewhat stable in-between models. Distance to regional urban centres is evidently more significant for predicting taxable earned income rather than tertiary education – it is however a positive trend stating that increased distance to regional urban centres increases the the distribution of high incomes. Distance to nearest amenity is highly significant for model (2) – (4) in relationship to tertiary education, but only significant at P = ≤ 0,05* in model (4) for taxable earned income, stating that it is the least significant variable in that model. Distance to nearest bus stop is highly significant in model (3) – (4) but only in model (4) for taxable earned income. Overall, the distance to railways proves itself to be the strongest predictor for both dependent variables. An interesting detail is that the R2- values have decreased due to the exclusion of the independent variable. For tertiary education, each model determines the strength of the regression to be circa 28 – 30 % and as low as circa 7 – 8 % for taxable earned income. In comparison to the R2-values provided when including the variable Distance to the CBD, this is a dramatic decrease. As for diagnostics, the VIF-values have also slightly decreased and are all < 2 for each model. The Durbin-Watson however, still indicates that there is auto-correlation and thereby a lack of independence for the estimations (see APPENDIX).

5.5. Maps over the residuals’ spatial distribution

After the regression analysis was performed, the data was used to create maps over the regressed variables’ spatial distribution in the Stockholm county. Map 9 is illustrating the residuals for the independent variables and tertiary education. Map 10 is illustrating the residuals for taxable earned income. Map 11 and 12 are illustrating the same correlations respectively when excluding distance to the CBD. The maps were produced using the software ArcMap since QGIS does not offer tools for this command. One of the benefits of using spatial regression when working with geographical data is the possibility of creating maps over the residuals in order to provide with a cartographical outline of how well the linear regression model is predicting the correlation. Each map presents a legend consisting of five classes that explains that the share of tertiary education and taxable earned incomes among the 400 nearest individuals that have lower values than what the regression estimated are displayed in blue plots on the map, and the ones with higher values are displayed in red. As evident in Map 9 and 10, the linear regression model predicts the relationship between the independent and the dependent variables well through the county, with the extreme values of the high shares of tertiary education and taxable earned income being concentrated in or near the CBD, with a tilt towards the north of the CBD. There are some values of extremely low shares of tertiary education and income in the south of the CBD, even if these areas are somewhat close to the CBD. When excluding variable X1 (Map 11 and 12), the geographical pattern remains quite stable but the linear model’s possibility of predicting the effect is significantly weaker, especially for taxable earned income.

38

Map 9 & 10: the residuals of the regressed variables’ spatial distribution in the Stockholm county

Map 11 & 12: the residuals of the regressed variables spatial distribution in the Stockholm county when excluding distance to the CBD

5.6. Hypothesis testing

The study was based on a theoretical framework of spatial capital that was building a link between how urban environments, depending on placement and mobility options could explain the geographical distribution of human capital and high incomes which have become central factors in the progression towards for sustainable cities. Five hypotheses based on that framework were formulated in order be be tested for empirical testing, scientifically evaluated and analysed. The thesis followed a pragmatic approach to whether a hypothesis should be

39 verified or falsified, initiating that verification would only confirm the theoretical model on middle-range level and falsification would set up one or several new hypotheses to tested in future experiments. These hypotheses were:

H1: “Shorter distances to the city business district and regional urban centres will both have the strongest effects on tertiary education and taxable earned income” • H0: “Distances to the city business district and regional urban centres will not have the strongest effect on the share of tertiary education nor taxable earned income” H2 “Shorter distance to the city business district will predict tertiary education and earned taxable income stronger then any other factor” • H0: “Distances to the city business district will not be the strongest variable” H3: “Shorter distances to railways, bus stops and amenities will correlate with a higher share of tertiary education and earned taxable income” • H0: “Neither distance will predict tertiary education and taxable earned income” H4: “Distances to railways will predict tertiary education and taxable earned income better than distances to bus stops” • H0: “Distance to railways and bus stops will not predict tertiary education and taxable earned income” H5: “Distances to amenities will have the weakest effect on tertiary education and taxable earned income” • H0: “Distance to amenities will not have a weaker effect than the other independent variables”

According to the findings in this study the initial results from the hypothesis testing were the following:

H1 has been falsified and the null-hypothesis can not be rejected. Shorter distances to the CBD predicted both tertiary education and taxable earned income better than distance to regional urban centres. In the case of the latter, further distance was proven to correlate positively with income, suggesting that households further away from the regional centres will have higher incomes and higher education levels.

H2 has been verified and the null-hypothesis can be rejected. Distance to the CBD predicted both tertiary education and taxable earned income better than any other variable in the analysis, with R2-values of ≈ 50 % for tertiary education and ≈ 20 % for taxable earned income.

H3 has been verified in regards to tertiary education when excluding the geographical proximity to central Stockholm, and the null-hypothesis can be rejected. The effects on taxable earned income were sporadic.

H4: has been verified and the null-hypothesis can be rejected. Distance to nearest railway was proven to be the strongest predictor after distance to the CBD and were statistically significant in all regression models except for model (4) for taxable earned income when not including distance to the CBD. This is believed to be a spurious correlation due to the effects from the other independent variables.

H5: has been falsified and the null-hypothesis can not be rejected. Nearest distance to amenity was the second strongest predictor to explain the distribution of tertiary education when excluding distance to the CBD but the weakest for predicting the distribution of taxable earned income.

40 5.7. Results summary

The results from the proximity analysis conducted through the use of the Distance matrix function in QGIS demonstrated that both tertiary education as well as taxable earned income featured strong geographical patterns of its distribution in the Stockholm county. In order to minimize the distances and avoid statistical outliers that could distort the normal distribution of the data, all the empirical material for the independent variables were modified by using the natural logarithm (Ln) for each observed distance and were also standardized. The multivariate regression analysis was then performed by using the five independent variables distance to the CBD, distance to regional urban centres, distance to nearest amenity, distance to nearest railway, and distance to nearest bus stop and had all statistically significant effects on the dependent variables tertiary education and taxable earned income. Like expected, independent variable distance to the CBD was proven to be the strongest explanatory factor for both the dependent variables, with R2-values around 50 % and 20 % respectively and was therefore excluded from the second analysis. When excluding the geographical proximity to the CBD, distance to nearest railway had the strongest effect on both tertiary education and taxable earned income, demonstrating a negative correlation. The remaining independent variables were proven to have more sporadic effects, although all statistically significant at P = ≤ 0,000*** except for distance to nearest amenity which was significant at P = ≤ 0,05* in model (5) for taxable earned income. The R2-values decreased dramatically when excluding distance to the CBD which indicates that the variance of the dependent variable that could be explained by the variance of the independent variable are rather weak. It is evident that the geographical proximity to central Stockholm has the strongest effect on the distribution of individuals with tertiary education and taxable earned income among the 400 nearest in the Stockholm county. The results also indicated that the spatial dimensions analysed in this study explains the distribution of tertiary education better than taxable earned income. There was however a problem regarding the auto-correlation being a strong factor in all cases, regardless of excluding the geographical proximity to the CBD. This means that the overall data analysed has a chance of being biased since the variables lack independence in their predictions.

6.0. Discussion

This chapter discusses the outcome of the study and what the empirical evidence suggests in regards to the theoretical framework that the study is based on. The first part of the chapter discusses the methodological approach in terms of its validity and robustness, strengths and weaknesses of the research design and the concepts of verification and falsification of the tested hypotheses. The second part discuses the theoretical concept of spatial capital and how it relates to empirical findings of Stockholm county as well as the field of urban and regional planning, both as a scientific discipline and as a professional practice.

6.1. Discussion on results and methodology

The results from the proximity analysis displayed a strong geographical concentration of individuals with tertiary education and individuals with high incomes in and near the central business district of Stockholm. In 2012, approximately 58 – 89 % of the inhabitants age 25 – 64 of that geographical location had received tertiary education, where as 27 – 60 % in the same year and in the same age had a taxable earned income in the highest decile (see Maps 5 & 6 in APPENDIX). The analysis was performed by the distance matrix function in QGIS and contributed with both cartographical and statistical data. It is evident that the distribution of tertiary education has a higher geographical density than taxable earned income, which is giving

41 support to Florida’s (2002) theory on how central urban environments are being occupied by the innovative economies, as well as Glaeser’s claim that individuals that possess higher human capital are attracted to cities due their cultural and aspirations and diversity, making cities consumer-oriented (the operationalization of amenities in the variable X4 consists of distances to museums, cinemas and theatres). The regression analysis provided with statistically significant evidence for all variables used and thereby supports three out of five hypotheses, which suggests that the larger components of the theoretical framework describes the situation of Stockholm county. H1 was falsified when excluding distance to the CBD, and can be rejected for the sake of the null-hypothesis; Distance to regional urban centres was not able to predict the share of individuals with tertiary education and high incomes to the same extent as distance to the CBD. This is an interesting notion because the regional urban centres should in theory attract more economically vital individuals because they are functioning as urban nodes for entrepreneurship and consumption. H5 was falsified in regards to tertiary education but not taxable earned income, meaning that it unexpectedly had a significant effect on the distribution of education and H2 and H4 did not. Distance to nearest amenity still had the weakest effect on taxable earned income. According to the spatial capital framework that differentiate between position and situation capital, both CBD and urban regional centres belong to the former category where economic vitality and knowledge (in a knowledge-based economy these are interrelated) should be flourishing. Shorter distances to the regional urban centres had no influence in model (2) and (3) for tertiary education and in model (1) for taxable earned income. It could be the that regional urban centres lack the amenities that attract the individuals with high human capital, which could be a possible investment strategy for increasing the human capital in the regional urban centres in the future development towards polycentrism emphasised in RUFS.

There is seemingly a discrepancy between the position capital variables which is conflicting with the theoretical claim of position vs. situation capital. According to the theory, the latter should compensate for the former, and the regional urban centres should compensate for the CBD by offering similar environments. This was not the case when excluding distance to the CBD. This might indicate that the spatial capital theory does not fully apply to the context of Stockholm county where a lot of the human capital and income are concentrated very near the CBD (with some expectations) even if the regional urban centres has a generally strong development of situation capital. Distance to railways were proven to be the strongest explanatory factor after distance to CBD, which was expected since railways are usually attractive means of public transport and areas that contain railway stations are prioritised for sustainable development according to RUFS. Considering that the the auto-correlation increased when excluding the CBD, it is possible that the correlation is spurious, when most railway stations are located in and near the CBD – but when examining Map 5 and 6, the values for both tertiary education and taxable earned income are all increasing when being near a railway station, even if the location is further away which indicates that there is a negative correlation.

Distance to nearest amenity had a somewhat strong effect on tertiary education, but weaker for taxable earned income when excluding distance to CBD, which again shines a light on the potential connection between possessing high human capital and being interested in cultural venues. Distance to bus stops had more sporadic effects which is believed being caused by the fact that there are much more of them in and within the county, making the proximity analysis somewhat difficult to interpret when far from all bus stops are located in or near the CBD or the regional urban centres, whilst almost all railways are. Overall, the analysis displays quite strong robustness in with the exception of the Durbin-Watson test. In regards to the R2-values

42 that decreased dramatically after excluding distance to CBD, especially for taxable earned income being as low as 7 – 8 %. This leaves no other conclusion that distance to the inner-city is what conceptualize most of Stockholm’s spatial capital and if this distance is lower, the higher is the share of the the human capital and high incomes, with some exceptions. However, the Durbin-Watson command displayed values < 1 which implies strong positive auto-correlation which means that the variable independence is violated and the analysis has over predicament when the residuals overlap in space, and also makes it difficult to draw any conclusions on causal inference.

It is also of importance to emphasise that the variable for taxable earned income is not intended as a proxy measurement of individual wealth, since the concept wealth can dependent on other factors than income, such as company net profits, property values, real-estate speculations and the stock market et cetera. It also important to emphasise that achieving tertiary education does not by any means guarantee an individual to obtain a taxable earned income in the highest decile. However, it is likely to assume that there is a significant share of individuals with high incomes that has a tertiary education and that this is expressed in spatial terms, which has been evident in this thesis findings.

6.2. Exploring Stockholm’s spatial capital

The question remains whether the theoretical framework for the spatial capital concept is sufficient enough to explain why the human capital and income levels geographical distribution in Stockholm county looks like it does. According to the theoretical conceptualization, the situation capital can compensate for the position capital by offering opportunities for spatial mobility. In the case of Stockholm, distance to the inner-city explains the distribution of the human capital and high incomes significantly better than the distance to regional urban centres, cultural venues, and public transportation. So the applicability of the theoretical framework to Stockholm county is a bit complex; all variables are proxy measurements based on the rationale of the theoretical implications, and there could be other variables that could have been included in the study besides the ones used.

All variables were proven to have significant effects on the geographical distributions, and when excluding distance to the CBD, the best predictor of high shares of tertiary education and taxable earned income was shorter distances to railways – meaning that if one lives for away from the CBD but close to a railway station that passes central Stockholm, the resident will likely have higher income and tertiary education than residents far away. So closeness to railways is the best compensator for not living near or in central Stockholm. The study does not exemplify what type of rail traffic that predicts the correlation the strongest, but it is reasonable to assume that metro stations predicts the relationship stronger than for instance commuter stations due to the fact that most metro stations already exists within the closeness to the CBD. The concept of causal inference, meaning if the there are any substantial evidence that the spatial capital actually causes the human capital’s and the incomes’ spatial distributions is still a methodological issue. The correlations do not imply causations, but they do imply that there is an effect and that some places in the county are more likely to benefit in human capital and income levels due the fact that certain distances are shorter and enables mobility.

So why is there such a strong concentration around the central Stockholm and what does that say in regards to urban sustainability? It has been established in Chapter 3 that central urban areas are optimal for economic growth in a knowledge-based economy, and that this requires a high human capital. Stockholm does not deviate from this claim but it can be other factors that

43 influence the population distribution; for instance, aspects of the physical geography, like the the fact that the CBD consists of several islands and that the county have had a sprawling suburbanization could make the central locations more ‘distinct’ and harder to densify and expand with buildings and infrastructure. This could mean that the urban morphological concepts that Oliveira (2016) and Marcus (2017) addresses occupies a smaller geography than most metropolitan cities do. If these qualities are considered attractive, the logical consequence is that they also become socially and economically exclusive. It could also be that the municipal planning monopoly that plays the key-role the planning process (see Chapter 2) and creates potential political barriers on how the city should be designed and connected. If a distance to railways increases the spatial (and presumably the social) mobility, and a municipality is not interested in using their lands for such expansions, social segregation is not surprising outcome. In regards to the sustainability approaches mentioned in RUFS, Stockholm’s regional development framework, the interest for creating social cohesions with the help of physical (spatial) planning is mentioned as well as maintaining the economic growth and minimizing environmental damages. It also suggests that areas that already have well-developed infrastructural and mobility options are prioritized in regards to urban and population densification (RUFS, 2016: p. 9 & 12). The development framework does emphasise that only 3-4 % of the residents and workplaces in the region’s rural parts and the archipelago has access to high-speed trains (ibid: p. 16). The framework does also emphasise the need to strengthen inter-municipal relations in to region in order to promote synergies related to sustainable development, but offers no practical solutions for overcoming political obstacles. It also emphasises the need of strengthening the regional urban cores with public transport, housing, workplaces, healthcare, culture and other social functions (ibid p. 9). This would mean that Stockholm is moving towards more polycentric functions where an increase of the human capital would be inevitable. But according to the the empirical findings in this thesis, distances to the regional urban cores does not explain human capital and high incomes to the same degree as distance to central Stockholm and railways. It could be in order for the regional planning process to emphasise the development of public transportation by railways as a top priority in order to overcome logistic boundaries and be able to utilize the spatial capital of the region, thus promoting the chances for achieving social cohesions, minimizing the individual car-use as well as social segregation.

The findings of the thesis suggest that the spatial capital theory is an efficient theoretical tool for urban planners to analyse the human and economic values of land, using Stockholm as example. It is however a rather ambiguous theory with different disciplinary approaches. It is therefore important to highlight Merton’s concept that addresses the understanding that elaborated theories of the magnitude should be met with the intention of be proven on a middle- range level, making room for hypothesis testing. By being able falsify the two of the four hypotheses (in this case the ones that suggested that distances to the regional urban centres would predict the dependent variables as well as the distance to the CBD, and that amenities would have the weakest effect) means that an opportunity of investigation of why these hypotheses were incorrect. The quantitative approach of combining GIS and regression analysis is a powerful method for analysing spatial data for this purpose, but the spatial capital theory does not exclude qualitative designs or other quantitative methods than used in this thesis.

7.0. Conclusion

The aim of this thesis was to investigate geographical proximities’ impact on the distribution of human capital and high incomes in Stockholm county. The study was conducted using the theoretical concept of spatial capital as a deductive framework that explains the human values

44 of land, as well as the concept of human capital to explain the impact of education and specialisation on economic growth. Spatial capital was divided into two binary sub-categories; position capital (physical location) and situation capital (mobility alternatives) The aim was also to analyse whether a spatial/human capital-link is theoretically relevant for urban sustainability and regional planning practices. The study was conducted through the use of a GIS-based proximity analysis, and multiple linear regression analysis using proxy data operationalized from the rationale of theoretical framework. The research questions and the answer were the following:

I. Are the distances to central Stockholm, the regional urban centres, cultural amenities as well as public transportation correlating with the distribution of the human capital as well as the incomes of individuals living in Stockholm county? II. Are the spatial distributions of human capital and incomes recognized or critically discussed in the theoretical understanding for social sustainability? If so, how does it relate to the regional planning process for Stockholm county?

Question I Yes, distances to the central Stockholm, regional urban centres, railways, bust stops, and amenities were all correlating negatively with the distribution of the human capital and income levels in the Stockholm county. Distance to the central business district (CBD) had the strongest effect of the human capital and income distribution. When excluding this variable; distance to railways had the strongest effect, and the remaining independent variables had sporadic effects between regression models, with distance to nearest amenity displaying the weakest effect on both income tertiary education but stronger effects on education. Overall, the regression models predicted tertiary education better than taxable earned income and displayed higher R2-values.

Question II Yes, human capital, especially in relation to social capital, are recognised by prominent urban theorist such as Richard Florida (2002, 2006), Edward Glaeser et al., (1992, 2004) and more, as a fundamental concept for achieving sustainable cities. There is no strict scientific definition of what social sustainability conceptualizes, but several scholars highlight the promotion of social equity and minimizing social disparity as key factors (see Colantonio, 2011; and Åhman, 2013). In a Swedish context, emphasis is put on minimizing social segregation and social exclusion which has become socio-political problem related to specific suburban areas (see Stigendal, 2012; and Sanandaji, 2014). The results from this study implies that the human capital and high income levels are highly concentrated to the CBD and nearby surroundings. This in turn indicates that central urban areas play a key role in social sustainability which is also highlighted in Stockholm regional development framework, RUFS (2016). It addresses the desire to maintaining economic growth and simultaneously prevent further environmental damage. It also addresses the need to identify how physical (spatial) planning can create a platform for social cohesions and equality, the need for housing densification near public transport, as well strengthening inter-municipal relationships to overcome political barriers and promote housing alternatives for those with weaker economic assets (p. 9, 12ff). However, the regional framework put little emphasis on the development of new public transportation alternatives or more cultural venues to strengthen the outskirts of the region, like for instance expansions of railways, which is likely to be the most attractive transportation alternative.

45 Final thoughts

The conclusion, based on the empirical findings is that the spatial capital concept provides with an understanding of how geographical proximities matter for the distribution of Stockholm county’s human capital and high incomes – which are central concepts in the work towards more sustainable cities, both in regards to environmental and economic factors as well as social. Using Stockholm as an example, the results demonstrated that shorter distances to both the position and the situation capital had statistically significant effects on higher shares of tertiary education and high incomes. Shorter distances to central Stockholm is believed to be the explanatory factor of the concentration of high shares of human capital and income because urban environments in the global north are characterized by knowledge-based (or ‘sustainable’) economies that requires specialisations that in turn generates wealth (Glaeser et al., 1992, 2004). In the case of Stockholm county, the public transportation alternatives can compensate for longer distances to the CBD in some cases, especially if it concerns public transportation by railways. RUFS addresses the ambition to develop the region through densification in urban environments that already are relatively well-developed in their infrastructure and transportation alternatives (RUFS, 2016: p. 18) which in turn means that these strategies are more oriented towards the developing places that already has a high spatial capital, but expresses very little about infrastructural expansions in order increase qualities that are associated with sustainability (in this case human capital and income) to places where the spatial capital is low. Considering the economic need for specialisation through education, as well as social segregation being a current problem in Sweden – it could be the case that the political arrangement of municipal self-governance in land regulation can be conflicting with the regional development framework’s ambitions to “connect the region” and enable polycentrism.

For future research, the spatial capital theory can be used to analyse the relationship to human capital and the high incomes in small to medium-sized cities in Sweden that might not demonstrate such stark spatial differences between the central city core and the suburban areas. It would also be interesting to study other Swedish regions that contains several cities and analyse how they relate to the position vs. situation capital model. Future research could also examine whether to include distance to nearest university or college in order to analyse whether it has an effect on tertiary education.

References

Literature

Aldrich, J. (1995). “Correlations genuine and spurious in Pearson and Yule” in Statistical science, pp. 364-376.

Andersson, Eva K., & Malmberg, Bo. (2016). “Segregation and the effects of adolescent residential context on poverty risks and early income career: A study of the Swedish 1980 cohort”. Urban Studies.

Barthon, C. & Monfroy, B (2010). “Sociospatial schooling practices: A spatial capital approach” in Educational Research and Evaluation 16(2): pp. 177–196.

Becker, G.S (1964). Human capital: a theoretical and empirical analysis, with special

46 reference to education. New York: Columbia University Press.

Blücher, G., & Graninger, G. (2006). Planering med nya förutsättningar: Ny lagstiftning, nya värderingar. Linköping Universisty Electronic Press.

Boudon, R (1991). “What middle-range theories are" in Contemporary Sociology (American Sociological Association) 20 (4), pp. 519-522.

Bourdieu, P (1985). “The social space and the genesis of groups”. Theory and Society 14(6), pp. 723–744.

Bradley, K. Hult, A. & Cars, G. (2013). “From Eco-Modernizing to Political Ecologizing: Future Challenges for the Green Capital” in Metzger, J & Olsson, A. (eds.) Sustainable Stockholm: Exploring Urban Sustainability in Europe’s Greenest City. New York: Routledge. pp. 168-194

Bryman, A. (2011). Samhällsvetenskapliga metoder. (2., [rev.] uppl.) Malmö: Liber.

Camagni, R. & Capello, R (2013). “Regional competitiveness and territorial capital: A conceptual approach and empirical evidence from the European Union” in Regional Studies, 47, pp. 1383–1402.

Chan, Y.H. (2004). “Biostatistics 201: linear regression analysis”. Singapore medical journal, 45 (2), pp. 55-61.

Coffman, C., & Gregson, M. E. (1998). Railroad development and land value. The Journal of Real Estate Finance and Economics, 16(2), pp. 191-204.

Colantonio, A. (2011). “Social Sustainability: Exploring the Linkages Between Research, Policy and Practice” in Jaeger, C. C., Tábara, D. & Jaeger, J. (eds.) Transformative Science Approaches for Sustainability. Berlin: Springer.

Cook, I. R. & Swyngedouw, E. (2012). “Cities, Social Cohesion and the Environment: Towards a Future Research Agenda”. Urban Studies, 49, pp. 1959-1979.

Cozens, P. M. (2008). “New Urbanism, Crime and the Suburbs: A Review of the Evidence”. Urban Policy and Research. Vol. 26: s. 429-444.

Conzen MRG (1960). “Alnwick Northumberland: a study in town-plan analysis”. Institute of British Geographers Publication 27. George Philip, London.

Dempsey, N., Bramley, G., Power, S. & Brown, C (2011). ””The social dimension of sustainable development: Defining urban social sustainability” in Sustainable Development, 19, pp. 289-300.

Dooling, S. (2009). “Ecological gentrification: A research agenda exploring justice in the city” in International Journal of Urban and Regional Research, 33(3), pp. 621-639.

Djurfeldt, G., Larsson, R. & Stjärnhagen, O. (2010). Statistisk verktygslåda 1: samhällsvetenskaplig orsaksanalys med kvantitativa metoder. (2. uppl.) : Studentlitteratur.

47 Florida, R. (2002). The rise of the creative class: and how it's transforming work, leisure, community and everyday life. New York: Basic Books

Florida, R. (2006). Cities and the Creative Class. Routledge

Forsberg, S. (2017). “The right to immobility’and the uneven distribution of spatial capital: negotiating youth transitions in northern Sweden”in Social & Cultural Geography, pp. 1-21.

Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2000). Quantitative geography: perspectives on spatial data analysis. Sage.

Frenkel, A., & Porat, I. (2017). “An integrative spatial capital-based model for strategic local planning–An Israeli case” in Planning Practice & Research, 32(2), pp. 171-196.

Friedmann, J. (1998). Planning theory revisited, European Planning Studies, 6 (3), pp. 245– 253.

Glaeser, E. L., Kallal, H. D., Scheinkman, J. A., & Shleifer, A. (1992). “Growth in cities”. Journal of political economy, 100(6), pp. 1126-1152.

Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2004). “Do institutions cause growth?” in Journal of economic Growth, 9(3), pp. 271-303.

Gunnarsson-Östling, U. Edvardsson Björnberg, K. & Finnvede, G. (2010). “Using the Concept of Sustainability: Interpretations in Academia, Policy and Planning” in Metzger, J & Olsson, A. (eds.) Sustainable Stockholm: Exploring Urban Sustainability in Europe’s Greenest City. New York: Routledge. pp. 51-70

Hillier B., & Hanson J. (1984). The social logic of space. Cambridge: Cambridge University Press,

Huang, X., Yang, Y., & Liu, Y.“Spatial capital or cultural capital? The residential choice of gentrifiers in Xuanwumen, Beijing” in Journal of Housing and the Built Environment, pp. 1- 19.

Jacobs J (1961). The death and life of great American cities. New York: Random house

Keil, R. (2007). “Sustaining Modernity, Modernizing Nature”. In: Krueger, R. & Gibbs, D. (eds.) Sustainable Development Paradox: Urban Political Economy in the United States and Europe. New York: Guilford Press.

Lévy J. (1994) L'espace légitime. Sur la dimension géographique de la fonc- tion politique, Fondation nationale des sciences politiques, Paris.

Lévy, J., & Lussault, M. (2003). Dictionnaire de la ge ́ographie et de l’espace des socie ́te ́s [Dictionnary of the geography and the space of society]. Belin: Paris.

Liljefors, P. (2016). Social sustainability in Swedish Urban Develeopment – what does it mean? A casestudy of three Citylab Action pilot projects. KTH, Skolan för arkitektur och samhällsbyggnad

Lynch, K. (1960). The image of the city. Cambridge: MIT press

48 Mace, A. (2017). “Spatial capital as a tool for planning practice” in Planning Theory, 16(2), pp. 119-132.

Marcus, L. (2008). “Spatial Capital and how to measure it” in New urbanism and beyond, pp. pp. 135-139.

Marcus, L. (2010). “Spatial capital” in The Journal of Space Syntax, Vol. 1(1), pp. 30-40.

Marcus, L., Balfors, B., Haas, T. (2013). “A sustainable urban fabric: the development and application of analytical urban design theory” in Metzger, J & Olsson, A (eds.) Sustainable Stockholm: exploring urban sustainability in Europé’s greenest city. London: Routledge, p. pp. 71-101.

Merton, R. K. (1949). On sociological theories of the middle range [1949]. na.

Metzger, J & Olsson, A. (2013). “Introduction: the greenest city?” in Metzger, J & Olsson, A. (eds.) Sustainable Stockholm: exploring urban sustainability in Europé’s greenest city. London: Routledge, pp. 1-9.

Newman, O. (1976). Defensible space: people and design in the violent city. London: Architectural Press.

Nielsen, M. M., Haandriksman, K., Christiansen, H., Costa, R., Sleutjes, B., Rogne, A., & Stonawski, M. (2017). Residential segregation in 5 European countries. Electronic resource.

Oliveira, V (2016). Urban morphology - an introduction to the study of the physical form of cities. Cham: Springer International Publishing.

Popper, K. (1959). The logic of scientific discovery. London: Hutchinson.

Putnam, R. D., Leonardi, R., & Nanetti, R. Y. (1994). “Making democracy work: Civic traditions in modern Italy” in Princeton university press.

Rauch, J. E. (1993). “Productivity gains from geographic concentration of human capital: evidence from the cities” in Journal of urban economics, 34(3), pp. 380-400.

Rérat, P., & Lees, L. (2011). “Spatial capital, gentrification and mobility: evidence from Swiss core cities” in Transactions of the Institute of British Geographers, 36(1), pp. 126-142.

Roy, A. & Ong, A. (eds.) (2011). Worlding Cities: Asian Experiments and the Art of Being Global, Oxford: Wiley-Blackwell.

Schmitt, P., Kahila, P., Östberg, S., & Stenström, J. (2006). Polycentrisk stadsutveckling ur ett nordiskt och europeiskt perspektiv: https://www.diva- portal.org/smash/get/diva2:700433/FULLTEXT01.pdf

Stigendal, M. (2012). “Segregation som blev utanförskap”. I&M. Invandrare & minoriteter, (1), pp. 5-9.

Stinchcombe, A.L. (2005). The logic of social research. Chicago: University of Chicago Press.

49 Storper, M., & Scott, A. J. (2009). “Rethinking human capital, creativity and urban growth” in Journal of economic geography, 9(2), pp. 147-167.

Swyngedouw, E. 2007. “Impossible Sustainability and the Postpolitical Condition”. In: Krueger, R. & Gibbs, D. (eds.) Sustainable Development Paradox: Urban Political Economy in the United States and Europe. New York: Guilford Press.

Trumberg, A. (2011). Den delade skolan: segregationsprocesser i det svenska skolsystemet (Doctoral dissertation, Örebro universitet)

Tufte, E.R. (2006). The Cognitive Style of PowerPoint: Pitching Out Corrupts Within (2nd ed.). Cheshire, Connecticut: Graphics Press.

Vallance, S., Perkins, H. C. & Dixon, J. E. (2011). “What is social sustainability? A clarification of concepts” in Geoforum, 42, pp. 342-348.

Voinov, A. (1998). Paradoxes of sustainability. Zhurnal Obshchei Biologii, 59, pp. 209-218.

Weingaertner, C., & Moberg, Å. (2014). “Exploring social sustainability: Learning from perspectives on urban development and companies and products” in Sustainable Development, 22(2), pp. 122-133.

Åhman, H. 2013. “Social sustainability – society at the intersection of development and maintenance” in Local Environment, 18, pp. 1153-1166.

Electronic sources

Boverket (2014). Plan- bygglagstiftelsens utveckling: https://www.boverket.se/sv/PBL- kunskapsbanken/Allmant-om-PBL/lag--ratt/plan--och-bygglagsstiftningens-utveckling/ [read 2018-05-03]

Boverket (2015). Detaljplaneprocessen: https://www.boverket.se/sv/PBL- kunskapsbanken/planering/detaljplan/detaljplaneprocessen/ [read 2018-05-03]

Boverket (2017). How Sweden is planned: https://www.boverket.se/en/start-in- english/planning/how-sweden-is-planned/ [read 2018-05-03]

Mäklarstatistik (2017). https://www.maklarstatistik.se/omrade/riket/stockholms- lan/stockholm/hagersten-liljeholmen/#/bostadsratter/arshistorik [read 2018-02-18]

Open Street Map (2016). Exportera: https://www.openstreetmap.org/export#map=14/19.8500/79.3515&layers=T [read 2018-03- 14]

Polisen. (2017). Nationella operativa avdelningen: https://polisen.se/Global/www%20och%20Intrapolis/Ovriga%20rapporter/Utsatta%20område n%20- %20social%20ordning,%20kriminell%20struktur%20och%20utmaningar%20för%20polisen. pdf [read 2018-05-25]

RUFS (2010). Regional Utvecklingplan för Stockholmsregionen:

50 http://www.rufs.se/publikationer/2016/samradsforslag/ [read 2018-04-12]

RUFS (2010). GIS-skikt: http://rufs.se/kartor/rufs-2010/gis-data-rufs-2010/gis-skikt/ [read 2018-02-23]

Sanandaji, T. (2014). Utanförskapets karta – en uppföljning av Folkpartiets rapportserie: http://www.dnv.se/nyheter/ny-rapport-utanforskapets-karta-en-uppfoljning-av-folkpartiets- rapportserie/ [read 2018-05-03]

SCB. (2018). Folkmängd i riket, län och kommuner 31 mars 2018 och befolkningsförändringar 1 januari–31 mars 2018: https://www.scb.se/hitta-statistik/statistik- efter-amne/befolkning/befolkningens-sammansattning/befolkningsstatistik/pong/tabell-och- diagram/kvartals--och-halvarsstatistik--kommun-lan-och-riket/kvartal-1-2018/ [read 2018-05- 03]

Sundell, Anders (2009). Guide: Regressionsanalys: https://spssakuten.wordpress.com/2009/12/21/regressionsanalys-1/

Vetenskapsrådet (2018). Forskningsetiska principer inom humanistisk-samhällsvetenskaplig forskning: http://www.codex.vr.se/texts/HSFR.pdf [read 2018-04-20]

51 APPENDIX

The APPENDIX contains larger resolutions of the GIS-based maps as well as more details on the parameter estimations from the regression analysis. This includes the tables; model summary, correlation matrix, coefficients and ANOVA for each multivariate analysis.

(Map 2: the regional urban centres of Stockholm county, Scale: 1: 250 000)

(Map 3: all bus stops in and to the Stockholm county. Scale: 1: 500 000)

52

(Map 5: all railway stops in the Stockholm county: Scale 1: 339 000)

(Map 6: all amenities in the Stockholm county: Scale 1: 330 000)

53

Map 7: geographical proximities between all dependent variables and tertiary education in. Scale 1: 160 000

Map 8: geographical proximities between all dependent variables and independent variable taxable earned income in blue Scale 1: 160 000

54

Maps 9 (above) and 10 over (below) over the residuals.

Map 9: Residuals for tertiary education

Map 10: Residuals for taxable earned income

55

Maps 11 (above) and 12 over (below) over the residuals minus distance to CBD

Map 9: Residuals for tertiary education

Map 10: Residuals for taxable earned income

56

Coefficients for Regression 1: Tertiary education

57

58 Coefficients for Regression 2: Taxable earned income

59

60 Coefficients for Regression 3: Tertiary education (- distance to CBD)

61

62 Coefficients for Regression 4: Taxable earned income (- distance to CBD)

63

64