- Email: [email protected]
Spatial restructuring and uneven intra-urban employment growth in metro- and non-metro-served areas in Copenhagen
Journal of Transport Geography 70 (2018) 21–30
Contents lists available at ScienceDirect
Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo
Spatial restructuring and uneven intra-urban employment growth in metroand non-metro-served areas in Copenhagen
T
⁎
Kristian Bothe , Høgni Kalsø Hansen, Lars Winther Department of Geosciences and Natural Resource Management, Geography Section, University of Copenhagen, Øster Voldgade 10, DK-1350 Copenhagen K, Denmark
A R T I C LE I N FO
A B S T R A C T
Keywords: Urban transport investments Employment growth Wider economic impacts Copenhagen Metro Intra-urban employment patterns
This paper addresses the wider benefits of major investments in urban transport and discusses the relevance of giving attention to time and geographical scale in the analysis of employment growth before, under and after the introduction of new urban transport infrastructure. Using descriptive statistics in combination with OLS modelling, the paper analyses the intra-urban employment growth by workplace in regard to the opening of the Copenhagen Metro in 2002. The study identifies strong employment growth in the case of Copenhagen and higher employment growth in metro-served areas compared to non-metro-served areas in the first ten years after the opening of the Metro. The study also finds that when zooming in on the local scale, employment growth has been unevenly distributed along the metro corridor leading to a spatial restructuring of intra-urban employment patterns. This highlights that geographical scale and time is of critical importance when addressing the development of employment in areas that have witnessed investment in infrastructure. Moreover, the paper shows that especially the existing urban structures of the built environment, supporting planning policies, the local economic context and the preconditions for the development seems to be of great importance when assessing intra-urban restructuring of employment.
1. Introduction Following the rise of the knowledge economy, urban competition and large-scale urban development projects from the early 1990s, the wider economic impacts of investments in transport have gained increasing political and academic attention (Banister and ThurstainGoodwin, 2011; Holvad and Leleur, 2015; Docherty and MacKinnon, 2013; Knowles and Ferbrache, 2016). Although diverse, theoretical and empirical findings support the hypothesis that especially agglomeration effects and labour market impacts can be substantial (Melo et al., 2013; Venables, 2007; Vickerman, 2007; Graham, 2007; SACTRA, 1999; Wangsness et al., 2016). This is also the case when addressing the geography of the labour market impacts. Improving internal transport accessibility does not just speed up connections that are already in place but potentially changes them and opens up new ones. Consequently, investments in urban transport infrastructure do not just support existing patterns and flows or trigger new developments, they also have a broader impact on a city's economic structures and flows (Scott, 2008; Graham and Marvin, 2001). Existing empirical research focusing on the direct impact of public transport investments on the labour market is limited in scope (Gibbons and Machin, 2006), and existing studies show rather divergent results, due to the different modes of transport that are
⁎
the focus of these studies, as well as different local contexts and supporting planning policies. Moreover, it is important to ask at what scale the impact is taking place (Beyazit, 2015; Meijers et al., 2012). Do infrastructural investments result in new economic activities, or do they have a more local re-distributional impact? From previous studies, we see divergent effects depending on the methodological approach and local context they adopt. Understanding the intra-urban impacts and the redistribution of economic activities is important because our current understanding of the wider economic impacts (e.g. agglomeration benefits and extensions of labour market catchment areas) are based on assumptions about the behavioural responses and focus on the aggregated regional or market scale (Banister and Thurstain-Goodwin, 2011; Vickerman, 2008). To address some of these potential shortcomings in the existing literature, the main objective of this paper is to scrutinize how intraurban employment structures change over time when a new large public transport investment is introduced. In this paper, we focus on the spatial restructuring of employment by workplace. We examine the intra-urban distribution of employment and local employment growth by considering the periods ten years before and after the introduction of the Copenhagen Metro in 2002. To highlight the methodological and
Corresponding author. E-mail addresses: [email protected] (K. Bothe), [email protected] (H.K. Hansen), [email protected] (L. Winther).
https://doi.org/10.1016/j.jtrangeo.2018.05.014 Received 18 August 2017; Received in revised form 12 May 2018; Accepted 15 May 2018 0966-6923/ © 2018 Elsevier Ltd. All rights reserved.
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Atlanta in the USA, for example, Nelson and Sanchez (1997) found that employment near MARTA stations rose significantly. Green and James (1993) have found a similarly significant positive effect of the Washington metro, whereas Cervero and Landis (1997) found more mixed employment effects in the case of BART in San Francisco and Bollinger and Ihlanfeldt (1997) that found neither a positive nor a negative employment impact from MATRA. In European cities, ex-post studies also show mixed effects on employment. In a comparative study, MejiaDorantes and Lucas (2014) found an increase in employment twice the rate of London across all station catchment areas in the first year after the opening of the Jubilee Line Extension. In Madrid, a positive employment effect in the catchment areas of the Metrosur was also identified. In both case studies, a divergent development was present, indicating that employment growth was not taking place in all transit served areas. Likewise, Mayer and Trevien (2015) found an increase in employment in municipalities near Paris served by the Regional Express Rail, whereas Padeiro (2013) found more mixed effects in a study of transport investments in suburban Paris. In the case of Istanbul, Beyazit (2015) found an increase in employment in areas served by the metro following its introduction, but a smaller increase than in surrounding areas of the city. Thus, despite an extensive theoretical and emperical literature, existing ex-post studies show rather divergent employment impacts. The mixed results can mainly be explained by the fact that different modes of transport were studied, as well as different local contexts and supporting planning policies. This indicates that context matters and that a range of necessary conditions needs to be in place to support a positive economic development (Banister and Berechman, 2001). Moreover, it may also serve as a reminder of the difficulties of isolating transport investments effects on the labour market. Accordingly, the present analysis do not aim at identifying direct effects of transport infrastructure investments on employment growth but rather point to how growth patterns differ between areas that have undergone transport investments and political attention versus areas that has not. In line with this, the scale of analysis can be considered to be of the utmost importance when examining the links between economic development and transport investments. Currently, the literature mainly focuses on the overall regional employment impact or changes in areas of geographical proximity to the investments in question, and in some cases control areas are used to document changes. There is, however, a risk that, by only focusing on parts of the urban area, the general impact of transport investments is concealed. Therefore, it is important to be aware of the extent to which employment growth in, for example, metro-served areas is a result of general employment growth in a region, increasing employment only in metro-served areas, or a result of a wider restructuring of employment on an intra-regional scale. Thus, to understand better the relation between investments in transport and employment growth, the local, regional and intra-regional outcomes all need to be analysed. In the remainder of this paper, the analysis and discussion will address the potential risk of overestimating the economic outcome of investment activities due to the relocation dynamics of intra-urban economic activity from, for example, non-metro-served areas to metroserved areas if only some parts of the relevant urban area are included in the analysis.
spatial challenges, employment restructuring is analysed in two ways. First, we assess the employment restructuring by summarizing and mapping employment growth on various intra-urban scales. Secondly, we analyse employment growth in metro- and non-metro-served areas using OLS regression models. This will provide empirical evidence of intra-urban restructuring of employment by workplace to illustrate the importance of the intra-urban scale. The models do not aim at estimating the direct effect of large infrastructural investments on employment growth within an urban area or seek to establish clear links between these investments and employment changes. Rather they serve to demonstrate how employment growth has developed differently within and outside metro served areas since the metro line in Copenhagen was decided and introduced. The remainder of the paper is structured as follows. In the following section, the links between investments in transport and the labour market are introduced in more detail, and earlier empirical findings are briefly reviewed. In the third section, the present case study of Copenhagen and the Copenhagen Metro and their context of development are presented. The following section introduces the data and the methodological approach chosen for the study. In the fifth section, intra-urban employment growth in Copenhagen between 1992 and 2012 is analysed, while the final section summarises and concludes the paper's findings. 2. Investments in urban transport and their links with the labour market As globalization has developed, it has been increasingly realized that transport is a critical determinant of both the performance of the urban economy and the attractiveness of the city as a place to live and work. Apart from the apparently positive effects of these investments, their wider impacts have gained increasing political and academic attention (Banister and Thurstain-Goodwin, 2011; Knowles and Ferbrache, 2016; Vickerman, 2007). Studies have highlighted that, under certain conditions, urban public transport investments in light rail and underground metro systems can act as catalysts for urban development and redevelopment (Gospodini, 2005). However, despite the development potential, only a few studies have addressed the impacts ex-post. In an extensive review, Gibbons and Machin (2006) stressed that existing research and studies specifically focusing on the impact on the labour market are limited in scope. According to Gibbons and Machin (2006), transport plays a three-way role in relation to labour markets, first, by affecting workers' behaviour and the labour supply; secondly, by affecting firms' decisions and the demand for labour; and thirdly, and as a result of these changes, by changing the equilibrium between supply and demand in the labour market. Two potential labour market impacts are widely acknowledged to occur when transport infrastructure is improved. The first relates to the overall expansion of the labour market catchment area. When transport accessibility is improved, workers are willing to commute longer distances, more people will enter the labour market, and jobs are potentially relocated to more accessible and higher productivity areas (Knowles and Ferbrache, 2016; SACTRA, 1999; Vickerman, 2007). Secondly, improvements in transport infrastructure influence spatial relationships, allowing potential agglomeration effects to occur (Graham, 2007; Venables, 2007; Melo et al., 2013). Agglomeration effects increase with the level of spatial proximity and concentration, and improving the transport network affects the level of concentration and density. This occurs either by increasing the effective density of an area by bringing workers and firms closer together or by relocating firms and/or workers from lower to higher productivity areas (Banister and Berechman, 2001; Wangsness et al., 2016). Regarding the effects on local employment, previous ex-post studies of investments in urban rail-based public transport have shown mixed results. Some studies find evidence of positive employment impacts. In
3. The case of Copenhagen and the Copenhagen Metro The Greater Copenhagen Region is the largest city-region in Denmark. In 2012 it had close to two million inhabitants, while the City of Copenhagen (the central municipalities) had just about 0.7 million inhabitants. Before the opening of the metro, public transport in the region was based on an extensive bus network and “S-bane” commuter train routes connecting the central city to five suburban “Finger Plan”corridors to the south west, west and north west, and two regional commuter train routes from Roskilde in the west and Helsingør in the 22
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
workplace addresses within 600 m walking distance were found (see Fig. 1). Earlier findings have shown that commercial development is most likely to be concentrated within 400 m of a transit station (Banister and Thurstain-Goodwin, 2011; Guerra et al., 2012), but also that the distance will vary from place to place depending on the local context. In this study, a walking distance of 600 m was chosen because urban development in the Greater Copenhagen area is restricted by planning principles near transit stations (Stationsnærhedsprincippet). Although the planning principles place restrictions on development, they allow more dense urban development within 600 m of transit stations. To test the walking distance of the catchment areas, larger catchment areas were also generated, but no significant employment effects were found beyond the 600-m limit. The metro-station catchment areas were divided into five spatially coherent areas based on their primarily urban function, local characteristics and the opening of the metro. The areas are referred to as metro-served areas (MSA), and the division is shown in Fig. 1. “MSA Inner City” is located in the old historical centre and is dominated by a high density of employment. “MSA Christianshavn” is located just south of the harbour and is a more mixed residential and commercial area characterized by large brown-field development sites near the harbour front. “MSA Frederiksberg” is mostly a residential area with a small number of small-scale brown-field sites. “MSA Ørestad” is a green-field development area with mixed land use. “MSA Amager East” is a diverse area dominated by mixed land use in the north, the airport in the south and large residential areas in between. Due to methodological problems in measuring the exact employment changes in the catchment area of the metro station in the airport, the station is not included in the analysis.1 It is though important to state that the airport has grown considerably in the studied period. In the analysis, the employment changes in the metro-served areas are compared to those in the nonmetro-served areas of the city. No specifically matched control areas are therefore used. This is because the metro-served areas cover a large part of the city, meaning that it was not possible to find comparable areas. This is especially the case for MSA Inner City and the green-field development in MSA Ørestad. National and regional trends in development are also included to benchmark the changes in employment. Ordinary Least Squares (OLS) models have been estimated to address intra-urban employment growth in the time period ten years after the metro was opened (2002−12). To test for potential differences in development trends before and after the financial crisis, OLS models have also been estimated for the first five years after the opening (2002–07) and the subsequent five years (2007–12). The main spatial level of analysis in the models is 500 × 500 m grid cells covering the entire City of Copenhagen. Grid cells, and not the metro-served areas described above, were chosen because on a small scale the existing administrative spatial divisions are outdated. Moreover, spatial variation in size in existing divisions makes it difficult to define the metroserved areas properly. Based on matched employee-employer data, all workplaces and numbers of employees have been aggregated for each grid cell. Due to the small number of workers in some peripherally placed cells, it was decided to merge cells with few employees with neighbouring cells. Following the merging exercise, all grid cells consist of at least 150 employees, with the City of Copenhagen being covered by 269 cells. The dependent variable is change in employment, defined as the absolute change in employment between 2002 and 2012. Absolute change has been chosen as the development variable instead of the employment growth rate because cells with few employees had a large influence and therefore caused problems. The main indicator we wish
north. The political decision to build the metro was taken with the Danish government's Ørestad Act of 1992. The investment marks a radical shift in Danish transport and economic development policies and contrasts greatly with policies in earlier decades that were dominated by a narrow focus on favouring car traffic and supporting equal economic development in the entire country. The key idea of the metro was to connect the less accessible southern part of the City of Copenhagen with the inner city and to develop a new large and attractive 310hectare business district, Ørestad, in a greenfield site owned by the city and the state on land reclaimed from the sea in the west of Amager. In 1996 the locations of the line and stations were approved for the first development phases of the metro, and construction started the same year. The metro consists of two lines in western and eastern Amager which are connected though the inner city and the western part of the City of Copenhagen (see Fig. 1). The metro system has 22 stations, of which nine are underground, and it is connected to regional or inter-city trains at four stations. The first phase connecting Ørestad to the inner city opened in 2002, and the second phase connecting the inner city with the western part of the city opened in 2003. The last phase connecting the airport opened in 2007. 4. Research methodology and data There is a contrast between the expected employment impacts of investments in urban public transport and the mixed employment impacts essentially found in ex-post studies. This paper address this issue by exploring the changing geographies of intra-urban employment ten years either side of the opening of the Copenhagen Metro. To assess local development we use employment growth by workplace as a proxy for intra-urban performance. The drawbacks of using employment as the dependent variable are that employment growth does not necessarily mirror the impact on productivity or the transformation in the industrial structure (Eriksson et al., 2017). However, gaining further knowledge of the spatial restructuring of employment growth by workplace is crucial because transport investments are often politically justified on their ability to generate jobs. The aim of the paper is to analyse intra-urban spatial employment structures within and between metro- and non-metro-served areas and to address the scale of the employment restructuring. Therefore, the study does not address the causality of the transport impacts or the direct effect of the metro investment. Employment growth in the period of the study is analysed in two steps. First, employment growth by workplace is addressed descriptively by summarizing them on different spatial scales and by mapping the intra-urban changes on a small scale using kernel densities estimations. Secondly, employment growth in metro- and non-metroserved areas is analysed through regression models. The analysis is based on individual register-based employment data with detailed information on workplace addresses in Copenhagen between 1992 and 2012. All data were obtained from Statistics Denmark, which holds detailed information on workplace addresses for all individuals. The data consist of all employed individuals in the age range of 16–65, and the high quality of the data on the registered workplaces makes it possible to analyse fully all changes in employment by workplace in Copenhagen in this period. The spatial scale is a key issue when analysing employment dynamics. The focus in this paper is on the local and intra-urban development, and the main study area is therefore the City of Copenhagen. This city is normally defined as the municipalities of Copenhagen and Frederiksberg (which is surrounded by Copenhagen), but in this analysis the neighbouring municipality of Tårnby is also included due to the layout of the metro. To address local employment growth, local catchment areas were identified and generated around each metro station, as none of the existing administrative intra-urban divisions of the city was found suitable for the analysis. Based on a detailed road and path network (the Danish Address and Road Database), all
1 During the period of the study, a large proportion of employment moves in and out of the 600-meter catchment area. This is mainly due to changes in the work addresses where people are registered. Also a large proportion of those who work geographically at the airport are registered at a different work address. Approximately 23,000 people are directly and indirectly employed at the airport.
23
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 1. The Copenhagen Metro and metro-served areas in Copenhagen.
impact. However, the controls were not significant, and they were therefore omitted from the final analysis. The goodness of fit tests was generally satisfactory but showed minor issues with heteroscedasticity and spatial autocorrelation in the models. The problem with heteroscedasticity is mainly due to there being few cells where large employment changes are taking place, especially in some metro-served areas, and we therefore decided not to treat them as outliers but to keep them in the models. Running the models with robust standard errors did not change the direction of the impacts but lowered the effect remarkably and therefore we chose to keep the models as OLS models.
to assess is the geographical placement of the cell in relation to the metro stations. Metro-served areas in the models are defined as cells where > 50% of employees' workplaces were located within 600 m of a metro station, and a dummy variable was used as a proxy for the spatial placement. Identifying the metro-served areas in this way causes a degree of uncertainty but was chosen in favour of the existing administrative spatial divisions for pragmatic reasons. To address the different opening years of the metro and the diverse influence in the different parts of the metro corridor, three dummy variables were constructed. The first dummy consists of all metro-served areas, the second of the metro-served areas that opened in 2002/2003 (excluding MSA Amager East), and finally five dummies were constructed for the five different parts of the metro corridor. The three models (A + B + C) are run for each time period using the dummies. All variables in the models are defined in further detail in Appendix A. In explaining changes in employment, we included several control variables in the model. First of all, we included earlier employment growth for 1992–2002, based on the assumption that the opening of the metro has changed the internal patterns of such growth. Secondly, we included the existing employment density in 2002 because the pre-existing distribution of jobs may influence where future jobs are created due to the presence of localization and urbanization economies. Thirdly, we also included a control for human capital defined as the share of employees in 2002 with at least a master's degree. The control is included because a number of studies have shown that a concentration of human capital is important for urban development because it generates spill-over effects and also triggers future concentrations of human capital (see e.g. Hansen and Winther, 2010 on Copenhagen). Initially we included controls for a mix of industries to test whether preexisting concentrations of specific industries such as KIBS had an
5. Intra-urban restructuring and employment growth In the following, employment growth ten years before and after the opening of the metro is analysed, first with a broad focus on the scale of the changes, and secondly with a narrower focus on the differences in growth patterns between and within metro-/non-metro-served areas ten years after the metro was opened. 5.1. Intra-urban employment growth Since the early 1990s, Copenhagen and the largest Danish cities have revitalised themselves around the knowledge and service economy (Hansen and Winther, 2012). This revitalisation is in great contrast to the 1970s and 1980s, which were dominated by a relocation of employment from the largest cities to small and medium-sized cities and from the central city to suburban municipalities in the Greater Copenhagen Area (Illeris, 1997; Winther, 2007; Hansen and Winther, 2010). From the early 1990s the direction of relocations changed, and during the 1990s and 2000s the four largest Danish cities had 24
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Table 1 Employment growth 1992–2012 in metro-served areas compared to changes in City of Copenhagen and Denmark. Employment in absolute numbers
Denmark Aarhus, Odense & Aalborg Greater Copenhagen Area City of Copenhagen - Non-metro-served areas - Metro-served areas - MSA Inner City - MSA Frederiksberg - MSA Christianshavn - MSA Ørestad - MSA Amager East
Change in employment in pct.
1992
1997
2002
2007
2012
92–97
97–02
02–07
07–12
92–02
02–12
2,558,205 329,284 880,548 340,488 248,191 92,297 50,441 22,371 9966 579 8940
2,649,855 346,631 918,723 347,176 257,432 89,744 47,825 22,197 10,137 828 8757
2,727,566 362,442 963,061 369,430 274,624 94,806 46,569 23,226 14,531 1601 8879
2,853,149 391,368 1,024,911 384,839 281,056 103,783 49,022 21,827 16,750 7788 8396
2,658,573 375,143 997,598 391,238 279,751 111,487 47,659 21,683 20,018 14,896 7231
3.58 5.27 4.33 1.96 3.72 −2.77 −5.19 −0.78 1.72 43.01 −2.05
2.93 4.56 4.82 6.41 6.68 5.64 −2.63 4.64 43.35 93.36 1.39
4.6 7.98 6.42 4.17 2.34 9.47 5.27 −6.02 15.27 386.45 −5.44
−6.82 −4.15 −2.69 1.66 −0.46 7.42 −2.78 −0.66 19.51 91.27 −13.88
6.62 10.07 9.37 8.5 10.65 2.72 −7.68 3.82 45.81 176.51 −0.68
−2.53 3.5 3.59 5.9 1.87 17.59 2.34 −6.64 37.76 830.42 −18.56
evident in recent years. Only 1600 people were employed in MSA Ørestad in 2002, but in 2012 the number was close to 15,000. In contrast to the slow start of Ørestad is the development of MSA Christianshavn. In these areas a large increase in employment is seen in the years before the opening of the metro and the last years of the period of the study. The situation here has to be understood in the context of the divergent development policies of the city government (Majoor, 2008) and the local context of the development. These areas are examples of brown-field development benefiting from the redevelopment of the harbour front. Until now, intra-urban employment growth has been described based on the aggregated catchment areas of the metro. To overcome the spatial constraints of using the defined metro-served areas and to address the broader changes in employment, employment densities and changes are mapped on a small scale for the entire city in Figs. 3, 4 and 5 using kernel density estimates based on employment data on 100 × 100 m grid cells. The maps show two important aspects of this intra-urban development. First of all, a clear concentration of employment is evident in the city centre in 2012, with smaller employment concentrations in a few specific areas in the city shown in Fig. 3. Secondly, a large intra-urban variation is seen in employment density change between 1992 and 2012 with both increasing and decreasing density trends (see Figs. 4 and 5). Between 1992 and 2002, a decreasing density is mostly seen in the centre whereas densification is seen in small pockets around the city and especially in some of the redeveloped harbour areas. The development around the metro stations also shows divergent trends. Where densification is seen around few metro stations, especially the metro station Christianshavn in MSA Christianshavn and metro station Frederiksberg in MSA Frederiksberg, a decreasing density is seen around the metro stations in the centre. In Fig. 5, a densification of employment between 2002 and 2012 is, apart from development around the National Hospital and the University Park in the northern part of the city, mostly seen in specific parts of the metro corridor and again in some of the redeveloped areas of the harbour front. This highlights two important aspects of the intra-urban employment development. Firstly, there seems to be a substantial spatial restructuring in the city in the period, and secondly there are large local variations within most of the MSA's. In MSA Christianshavn densification is seen around both metro stations, while no clear pattern and both an increasing and decreasing density is seen around the metro stations in MSA Frederiksberg and MSA Inner City. In MSA Ørestad, employment densification is mainly taking place around two of the metro stations. From the first part of the analysis, it can be seen that employment growth was higher in metro-served areas than in non-metro-served areas in the ten years after the opening. By addressing employment growth in the different metro-served areas, it is also clear that large local differences in employment growth took place and that by mapping them on a micro-scale, a mixed pattern of increasing and decreasing
employment growth above the national average. Whereas 26% of all national employment was located in the four largest Danish cities in 1992, this increased to 29% in 2012. In Table 1 employment growth in 1992–2012 is summarized ranging from the national and urban level down to the metro's catchment areas. The table shows that overall employment growth in the City of Copenhagen differs from trends at the national level and in the other three largest cities in Denmark (Aarhus, Odense and Aalborg) throughout the period. This is also seen in the period 2007–2012, when, despite the financial crisis, the City of Copenhagen had a small growth in employment, while national employment declined markedly. Focusing on intra-urban employment growth in the City of Copenhagen, Table 1 reveals that employment growth in the metroserved areas differs considerably from trends in the rest of the city. In the ten years after the political decision to build the metro was taken and until the opening of its first stage in 2002, employment rose by 2.6% in the metro-served areas, compared to employment growth of 10.7% in the remainder of the city. In the ten years after the metro opened, this picture changed, and the metro-served areas had dramatically greater employment growth (17.6%) compared to the rest of the city (1.9%). With more than three out of four new jobs located in the metro-served areas in this period, the redistribution of employment in Copenhagen has been considerable in this period. From 2002 to 2012 the level of employment in Copenhagen located in metro-served areas increased from 24.9% to 28.1%. Despite overall employment growth in the metro-served areas in 2002–2012, employment growth is unevenly distributed between the different areas. Fig. 2 displays annual growth in employment in the metro-served areas. It can be seen that, while employment grew in the western corridor south of the harbour (MSA Ørestad and MSA Christianshavn), employment levels in the other corridors was constant or decreased slightly throughout this period. The developments observed in MSA Amager East and MSA Frederiksberg were not unexpected because they mostly consist of already existing or newly developed residential areas. The decrease in employment around the metro stations in MSA Inner City is more surprising, and it contrasts with the development patterns seen in several earlier ex-post studies of metro and light rail investments from other European cities. This indicates that the implementation of the metro in Copenhagen do not coincide with boosted employment in the inner city but rather has opened up new economic spaces in the city, where new and redistributed developments are taking place. The levels of employment growth in MSA Ørestad and MSA Christianshavn are examples of this type of development. As already stressed, the development of Ørestad as a new modern business district was one of the main aims of investing in the metro. Despite a slow start in developing Ørestad, which was mostly driven by the relocation of the Danish Broadcasting Company and larger public investments in universities (Knowles, 2012), a significant increase in employment is 25
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 2. Yearly employment growth in 1992–2012 in metro-served areas in City of Copenhagen.
Fig. 3. Employment density in City of Copenhagen 2012. 26
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 4. Employment density change in City of Copenhagen 1992–2002.
growth, indicating that employment growth prior to 2002 influences the employment growth after 2002. Concerning employment growth between 2002 and 07 and 2007–12, we find the same trend in the two periods, but with a higher positive relationship between employment growth and the metroserved areas in the second period. Moreover, important differences are found between models 2C and 3C, where the focus is on different parts of the metro corridor. In the first five years from 2002 to 2007, a significant positive relationship between employment growth and MSA's are found in MSA Ørestad and MSA Inner City while a significant negative relationship is identified in MSA Amager East, where the metro opened later. From 2007 to 2012 a significant positive relationship is still seen in MSA Ørestad and MSA Christianshavn, but not in MSA Inner City. To check for spatial autocorrelation we use a Moran's I test. The test show a small but significant clustering of the residuals in four of the models (model 1A, 1B, 3A and 3C), indicating that the residuals are not randomly distributed. To address the autocorrelation, a Getis-Ord hotspot analysis (Gi* statistic) was done for each set of the residuals. Based on the analysis, a spatial concentration of high values was identified for each model in cells covering Ørestad (for all four models) and the inner harbour (model 3A and 3B). To control for these hot-spot-areas, four different dummies were included in the four models. By doing so, the spatial autocorrelation becomes insignificant, while the results remain robust. In general, the models demonstrate a positive relationship between employment growth and the metro-served areas in the case of Copenhagen. In particular, the metro-served areas in Ørestad and in the central parts of the city seem to have benefitted from employment
employment density is seen. This emphasizes the importance of scale when adressing employment restructuring in metro-served areas. Employment growth is not taking place in all metro-served areas, but primarily in the redeveloped areas south of the harbour and in the new town development in Ørestad. The metro-served areas in the inner city are losing employment relatively, as are some of the other parts of the metro corridor.
5.2. Employment growth in metro- and non-metro-served areas Table 2 presents the output of the regressions of three periods after the metro opened. Concerning employment growth in the first ten years after the opening, we find a positive relationship between geographically being located in a metro-served area and employment growth. The relationship is largest and most significant in models 1B and 1C, indicating that the timing of the opening and spatial variation inside the metro corridor are important for understanding the relationship. Model 1C highlights the earlier descriptive findings and underlines the fact that, while a positive significant relation is seen in the metro-served areas of MSA Ørestad and MSA Christianshavn, a significant negative relation is seen in the metro-served areas of MSA Amager East. Controlling for the level of human capital shows that the pre-existing level has a significant, positive, but limited contribution to employment growth. Controlling for employment density shows a negative but not highly significant relation in some of the models. This indicates that to some extent employment growth is taking place in new and lower density areas of the city. However, this should be seen in relation to the positive, but not greatly significant impact of earlier employment 27
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 5. Employment density change in City of Copenhagen 2002–2012.
employment structures in Copenhagen differ between areas served by a metro and areas that are not. This is done by using register data along with catchment areas based upon distance to metro stations. Several conclusions can be drawn from the analysis. First and foremost, the analysis shows that scale is of great importance when understanding and questioning the changes within and between metro- and nonmetro-served areas. By addressing the spatial restructuring on an aggregated level, employment growth was significantly higher in metroserved areas compared to non-metro-served areas in Copenhagen in the first ten years after the metro opened. Addressing the employment changes on a small scale also stressed that employment growth was unevenly distributed along the different parts of the metro corridor. Especially in the west of Amager, the introduction of the metro seems to occur simultaneously with employment growth in areas that, before the introduction of the metro, were poorly connected to the rest of the city. Despite significant improvements in transport accessibility in the inner city, the inner city lost employment relative to other areas in this period. The maps in Figs. 4 and 5 provided a visual impression of where increases in employment density have taken place within the City of Copenhagen. Employment density has mainly increased along the metro line and by the harbour front, although a few exceptions can be identified. Just as importantly, by zooming out from the metro line and observing the development across the city, considerable decreases can also be identified. This emphasizes the importance of scale when discussing the effects of urban investments. Secondly, Fig. 2 and Table 1, underlined by the various OLS models, demonstrate that not only is scale an important issue to keep in mind when assessing the impact of transport investments, but so is the time variable. The analysis shows that areas around the metro grew in
growth during the period in question. However, the models do not allow us to assign this positive relationship as an outcome of the introduction of the metro. Most likely, employment growth in metroserved areas is a combination of changing economic structures caters for new location dynamics, investments in the metro and a dedicated political focus on developing and redeveloping specific areas of the city. Also, the models suggest that some metro-served areas of the city, such as Amager East and Frederiksberg in the early period and the inner city in the later period, show negative relationships between employment growth and being located in a metro-served area. This highlights the important question of to what extent the restructuring of employment in the city is caused by a general restructuring of the economy catering for new industries and new location dynamics, the creation of new employment, the emergence of the metro, or by redistribution of already existing employment from one area to another. Lastly, the model also reflects what has already been documented in the case of Copenhagen (Hansen and Winther, 2010), namely that economic development is closely linked to the level of human capital. Here it is shown that, even when looking at small areas, such as metro catchment areas in Copenhagen, the picture is the same: an increase in employment opportunities is more likely to be found in places that already have a high level of human capital. This stresses the difficulties in isolating any direct effect of investments in the metro and emphases that employment trends in the period should rather be seen as a series of effects with different degrees and sorts of influence.
6. Conclusion The aim of this paper has been to examine how intra-urban spatial 28
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Table 2 OLS models on employment growth in the City of Copenhagen 2002–2012. 2002–12 Model 1A Earlier growth 1992–2002 Employment density 2002 Human Capital 2002 All MSA
⁎
1.522 (0.826) −65.534 (53.696) 14.634⁎⁎⁎ (4.968) 227.792⁎ (121.346)
2002–07 Model 1B ⁎⁎
1.790 (0.805) −98.500⁎ (52.696) 11.877⁎⁎ (4.880)
MSA Ørestad MSA Amager East MSA Frederiksberg MSA Inner City MSA Christianshavn
R2 Adj. R2 F-test Mean VIF Moran's I N
1.108 (0.724) −73.005 (48.551) 10.990⁎⁎ (4.528)
Model 2A 0.832 (0.513) −6.152 (33.356) 11.063⁎⁎⁎ (3.086) 52.126 (75.380)
620.614⁎⁎⁎ (142.544)
First MSA
Intercept
Model 1C
2007–12
255.599 (336.405) 0.060 0.046 4.198⁎⁎⁎ 1.12 0.021⁎ 269
473.296 (331.635) 0.111 0.098 8.247⁎⁎⁎ 1.15 0.024⁎⁎ 269
Model 2B ⁎
0.986 (0.506) −24.583 (33.110) 9.765⁎⁎⁎ (3.066)
Model 2C
Model 3A
Model 3B
Model 3C
0.741 (0.475) −30.243 (31.851) 9.338⁎⁎⁎ (2.971)
0.690 (0.568) −59.382 (36.953) 3.571 (3.419) 175.667⁎⁎ (83.509)
0.804 (0.562) −73.917⁎⁎ (36.777) 2.111 (3.406)
0.366 (0.530) −42.763 (35.543) 1.652 (3.315)
275.073⁎⁎⁎ (89.563) 2915.935⁎⁎⁎ (331.343) −466.888⁎⁎⁎ (171.295) −78.745 (183.464) 444.757⁎ (252.300) 642.481⁎⁎⁎ (245.721) 368.130 (305.372) 0.304 0.282 14.171⁎⁎⁎ 1.13 0.003 269
−67.068 (208.975) 0.070 0.056 4.972⁎⁎⁎ 1.12 0.012 269
46.037 (208.372) 0.101 0.087 7.374⁎⁎⁎ 1.15 0.016 269
345.541⁎⁎⁎ (99.482) 1273.538⁎⁎⁎ (217.374) −336.523⁎⁎⁎ (112.376) −105.880 (120.359) 504.979⁎⁎⁎ (165.518) 133.666 (161.202) 118.452 (200.336) 0.232 0.208 9.807⁎⁎⁎ 1.13 0.004 269
322.667 (231.511) 0.028 0.013 1.893⁎ 1.12 0.037⁎⁎⁎ 269
427.259⁎ (231.449) 0.055 0.040 3.825⁎⁎⁎ 1.15 0.034⁎⁎⁎ 269
1642.398⁎⁎⁎ (242.570) −130.364 (125.402) 27.135 (134.310) −60.222 (184.704) 508.816⁎⁎⁎ (179.887) 249.679 (223.557) 0.185 0.160 7.388⁎⁎⁎ 1.13 0.016 269
Standard errors in parentheses. ⁎ p < 0.10. ⁎⁎ p < 0.05. ⁎⁎⁎ p < 0.01.
employment in Ørestad, the initially expected development did not take place there due to a political shift in the period towards a more liberal and market-based planning regime that especially favoured development of the harbour front. Secondly, it is important to understand the local context and the preconditions for the development. Using the OLS models it was shown that employment growth took place especially in areas that already had a high level of human capital. Last but not least, an understanding of the location preferences of firms seems to be important. Although not addressed directly in the present analysis, employment growth is taking place especially within the service and knowledge-intensive sectors, and there are signs of changed location preferences favouring the harbour front and some of the metro-served areas. Therefore, understanding how the location preferences in these sectors are directly affected by investments in transport is an important task for future research.
different time periods, some taking off parallel with the construction of the metro, while others started to blossom at a later stage. This stresses the difficulties in assigning employment changes to an outcome of a large transport investment like the Copenhagen Metro and moreover suggests that results of ex-post studies are highly sensitive to when they are carried out. Thirdly, the interpretations of the outcome of this study support earlier findings showing that a range of necessary conditions need to be in place to stimulate economic development when large public transport investments are made (Banister and Berechman, 2001). In the case of Copenhagen especially three conditions seem to be of great importance when assessing changes to employment. The first relates to the availability of suitable land/space and supporting planning policies. In Copenhagen, employment growth is not random but is to a large degree limited by the existing urban structures of the built environment, taking place especially in brown-field and green-field development areas. Therefore, the development in employment structures needs to be understood in the context of how former industrial areas are becoming available with the transformation towards the service and knowledge economy. Closely related to this, employment changes also need to be seen in the context of supporting planning policies and how these policies change throughout the period. Despite a large increase in
Funding This work was supported by Metroselskabet (108567) as part of the PhD-project “The wider socio-economic and spatial impacts of urban transport investments”.
Appendix A. Variables used in the OLS regressions
Variables
Definition
Dependent variables Employment growth 2002–12 Employment growth 2002–07 Employment growth 2007–2012
Absolute employment change in 2002–12 Absolute employment change in 2002–07 Absolute employment change in 2007–12 29
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Metro served areasa All MSA First MSA MSA Ørestad MSA Inner City MSA Frederiksberg MSA Amager East MSA Chr/IB Controllers Earlier growth 1992–2002 Employment density 2002 Human Capital 2002
Dummy Dummy Dummy Dummy Dummy Dummy Dummy
variable variable variable variable variable variable variable
indicating indicating indicating indicating indicating indicating indicating
all metro served grid cells grid cells that where metro served in 2002/2003 the metro served grid cells in Ørestad the metro served grid cells in Inner city the metro served grid cells in Frederiksberg the metro served grid cells in Amager East the metro served grid cells in Christianshavn
Employment change (%) in 1992–2002 Total number of people employed in 2002 (log) Share (%) of workers in 2002 with at least a master's degree (ISCED 5A)
a A grid cell is defined as a metro-served grid cell if > 50% of the people employed in the cell have a work address located within 600 m walking distance of a metro station.
Development. Edward Elgar, pp. 259–272. Illeris, S., 1997. The changing location of service activities in the Copenhagen Region. Geografisk Tidsskrift-Danish J. Geogr. 1997, 120–142. Knowles, R.D., 2012. Transit oriented development in Copenhagen, Denmark: from the finger plan to Ørestad. J. Transp. Geogr. 22, 251–261. Knowles, R.D., Ferbrache, F., 2016. Evaluation of wider economic impacts of light rail investment on cities. J. Transp. Geogr. 54, 430–439. Majoor, S., 2008. Progressive planning ideals in a neo-liberal context, the case of Ørestad Copenhagen. Int. Plan. Stud. 13, 101–117. Mayer, T., Trevien, C., 2015. In: Papers, C.E.P.R.D. (Ed.), The Impacts of Urban Public Transportation: Evidence from the Paris Region, pp. 31 INSEE Working paper G2015/0. Meijers, E., Hoekstra, J., Leijten, M., Louw, E., Spaans, M., 2012. Connecting the periphery: distributive effects of new infrastructure. J. Transp. Geogr. 22, 187–198. Mejia-Dorantes, L., Lucas, K., 2014. Public transport investment and local regeneration: a comparison of London's Jubilee Line Extension and the Madrid Metrosur. Transp. Policy 35, 241–252. Melo, P.C., Graham, D.J., Brage-Ardao, R., 2013. The productivity of transport infrastructure investment: a meta-analysis of empirical evidence. Reg. Sci. Urban Econ. 43, 695–706. Nelson, A.C., Sanchez, T.L., 1997. The Influence of MARTA on Population and Employment Location. Paper Presented at the 76th Annual Meeting of the Transportation Research Board, Washington, DC. Padeiro, M., 2013. Transport infrastructures and employment growth in the Paris metropolitan margins. J. Transp. Geogr. 31, 44–53. SACTRA, 1999. Transport and the Economy - Final Report. Standing Advisory Committee on Trunk Road Appraisal, London. Scott, A.J., 2008. The Social Economy of the Metropolis. Oxford University Press, Oxford. Venables, A.J., 2007. Evaluating urban transport improvements: cost-benefit analysis in the presence of agglomeration and income taxation. J. Transp. Econ. Pol. 41, 173–188. Vickerman, R., 2007. Recent evolution of research into the wider economic benefits of transport infrastructure investments. In: JTRC Discussion Paper. 2007. Vickerman, R., 2008. Transit investment and economic development. Res. Transp. Econ. 23, 107–115. Wangsness, P.B., Rødseth, K.L., Hansen, W., 2016. A review of guidelines for including wider economic impacts in transport appraisal. Transp. Rev. 1–22. Winther, L., 2007. Location dynamics of business services in the urban landscape of Copenhagen: imaginary spaces of location. Belgeo 2007, 51–72.
References Banister, D., Berechman, Y., 2001. Transport investment and the promotion of economic growth. J. Transp. Geogr. 9, 209–218. Banister, D., Thurstain-Goodwin, M., 2011. Quantification of the non-transport benefits resulting from rail investment. J. Transp. Geogr. 19, 212–223. Beyazit, E., 2015. Are wider economic impacts of transport infrastructures always beneficial? Impacts of the Istanbul Metro on the generation of spatio-economic inequalities. J. Transp. Geogr. 45, 12–23. Bollinger, C.R., Ihlanfeldt, K.R., 1997. The impact of rapid rail transit on economic development: the case of Atlanta's MARTA. J. Urban Econ. 42, 179–204. Cervero, R., Landis, J., 1997. Twenty years of the Bay Area rapid transit system: land use and development impacts. Transp. Res. A 31, 309–333. Docherty, I., MacKinnon, D., 2013. Transport and economic development. In: Rodrigue, J.P., Notteboom, T., Shaw, J. (Eds.), The Sage Handbook of Transport Studies. Sage, London, UK, pp. 226–240. Eriksson, R.H., Hansen, H.K., Winther, L., 2017. Employment growth and regional development: industrial change and contextual differences between Denmark and Sweden. Eur. Plan. Stud. 25, 1756–1778. Gibbons, S. & Machin, S. 2006. Transport and Labour Market Linkages: Empirical evidence, implications for policy and scope for further research. Background paper for the Eddington report to the Department of Transport. Gospodini, A., 2005. Urban development, redevelopment and regeneration encouraged by transport infrastructure projects: the case study of 12 European cities. Eur. Plan. Stud. 13, 1083–1111. Graham, D.J., 2007. Agglomeration, productivity and transport investment. J. Transp. Econ. Pol. 41, 317–343. Graham, S., Marvin, S., 2001. Splintering Urbanism: Networked Infrastructures, Technological Mobilities and the Urban Condition. Routledge, London. Green, R.D., James, D.M., 1993. Rail Transit Station Area Development: Small Area Modeling in Washington, D.C. M.E. Sharpe, Armonk, NY. Guerra, E., Cervero, R., Tischler, D., 2012. Half-mile circle. Transp. Res. Rec. 2276, 101–109. Hansen, H.K., Winther, L., 2010. The spatial division of talent in city regions: location dynamics of business services in Copenhagen. Tijdschr. Econ. Soc. Geogr. 101, 55–72. Hansen, H.K., Winther, L., 2012. The Urban Turn: Cities, Talent and Knowledge in Denmark. Aarhus Universitetsforlag, Aarhus. Holvad, T., Leleur, S., 2015. Transport projects and wider economic impacts. In: Hickman, R., Givoni, M., Bonilla, D., Banister, D. (Eds.), Handbook on Transport and
30
Contents lists available at ScienceDirect
Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo
Spatial restructuring and uneven intra-urban employment growth in metroand non-metro-served areas in Copenhagen
T
⁎
Kristian Bothe , Høgni Kalsø Hansen, Lars Winther Department of Geosciences and Natural Resource Management, Geography Section, University of Copenhagen, Øster Voldgade 10, DK-1350 Copenhagen K, Denmark
A R T I C LE I N FO
A B S T R A C T
Keywords: Urban transport investments Employment growth Wider economic impacts Copenhagen Metro Intra-urban employment patterns
This paper addresses the wider benefits of major investments in urban transport and discusses the relevance of giving attention to time and geographical scale in the analysis of employment growth before, under and after the introduction of new urban transport infrastructure. Using descriptive statistics in combination with OLS modelling, the paper analyses the intra-urban employment growth by workplace in regard to the opening of the Copenhagen Metro in 2002. The study identifies strong employment growth in the case of Copenhagen and higher employment growth in metro-served areas compared to non-metro-served areas in the first ten years after the opening of the Metro. The study also finds that when zooming in on the local scale, employment growth has been unevenly distributed along the metro corridor leading to a spatial restructuring of intra-urban employment patterns. This highlights that geographical scale and time is of critical importance when addressing the development of employment in areas that have witnessed investment in infrastructure. Moreover, the paper shows that especially the existing urban structures of the built environment, supporting planning policies, the local economic context and the preconditions for the development seems to be of great importance when assessing intra-urban restructuring of employment.
1. Introduction Following the rise of the knowledge economy, urban competition and large-scale urban development projects from the early 1990s, the wider economic impacts of investments in transport have gained increasing political and academic attention (Banister and ThurstainGoodwin, 2011; Holvad and Leleur, 2015; Docherty and MacKinnon, 2013; Knowles and Ferbrache, 2016). Although diverse, theoretical and empirical findings support the hypothesis that especially agglomeration effects and labour market impacts can be substantial (Melo et al., 2013; Venables, 2007; Vickerman, 2007; Graham, 2007; SACTRA, 1999; Wangsness et al., 2016). This is also the case when addressing the geography of the labour market impacts. Improving internal transport accessibility does not just speed up connections that are already in place but potentially changes them and opens up new ones. Consequently, investments in urban transport infrastructure do not just support existing patterns and flows or trigger new developments, they also have a broader impact on a city's economic structures and flows (Scott, 2008; Graham and Marvin, 2001). Existing empirical research focusing on the direct impact of public transport investments on the labour market is limited in scope (Gibbons and Machin, 2006), and existing studies show rather divergent results, due to the different modes of transport that are
⁎
the focus of these studies, as well as different local contexts and supporting planning policies. Moreover, it is important to ask at what scale the impact is taking place (Beyazit, 2015; Meijers et al., 2012). Do infrastructural investments result in new economic activities, or do they have a more local re-distributional impact? From previous studies, we see divergent effects depending on the methodological approach and local context they adopt. Understanding the intra-urban impacts and the redistribution of economic activities is important because our current understanding of the wider economic impacts (e.g. agglomeration benefits and extensions of labour market catchment areas) are based on assumptions about the behavioural responses and focus on the aggregated regional or market scale (Banister and Thurstain-Goodwin, 2011; Vickerman, 2008). To address some of these potential shortcomings in the existing literature, the main objective of this paper is to scrutinize how intraurban employment structures change over time when a new large public transport investment is introduced. In this paper, we focus on the spatial restructuring of employment by workplace. We examine the intra-urban distribution of employment and local employment growth by considering the periods ten years before and after the introduction of the Copenhagen Metro in 2002. To highlight the methodological and
Corresponding author. E-mail addresses: [email protected] (K. Bothe), [email protected] (H.K. Hansen), [email protected] (L. Winther).
https://doi.org/10.1016/j.jtrangeo.2018.05.014 Received 18 August 2017; Received in revised form 12 May 2018; Accepted 15 May 2018 0966-6923/ © 2018 Elsevier Ltd. All rights reserved.
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Atlanta in the USA, for example, Nelson and Sanchez (1997) found that employment near MARTA stations rose significantly. Green and James (1993) have found a similarly significant positive effect of the Washington metro, whereas Cervero and Landis (1997) found more mixed employment effects in the case of BART in San Francisco and Bollinger and Ihlanfeldt (1997) that found neither a positive nor a negative employment impact from MATRA. In European cities, ex-post studies also show mixed effects on employment. In a comparative study, MejiaDorantes and Lucas (2014) found an increase in employment twice the rate of London across all station catchment areas in the first year after the opening of the Jubilee Line Extension. In Madrid, a positive employment effect in the catchment areas of the Metrosur was also identified. In both case studies, a divergent development was present, indicating that employment growth was not taking place in all transit served areas. Likewise, Mayer and Trevien (2015) found an increase in employment in municipalities near Paris served by the Regional Express Rail, whereas Padeiro (2013) found more mixed effects in a study of transport investments in suburban Paris. In the case of Istanbul, Beyazit (2015) found an increase in employment in areas served by the metro following its introduction, but a smaller increase than in surrounding areas of the city. Thus, despite an extensive theoretical and emperical literature, existing ex-post studies show rather divergent employment impacts. The mixed results can mainly be explained by the fact that different modes of transport were studied, as well as different local contexts and supporting planning policies. This indicates that context matters and that a range of necessary conditions needs to be in place to support a positive economic development (Banister and Berechman, 2001). Moreover, it may also serve as a reminder of the difficulties of isolating transport investments effects on the labour market. Accordingly, the present analysis do not aim at identifying direct effects of transport infrastructure investments on employment growth but rather point to how growth patterns differ between areas that have undergone transport investments and political attention versus areas that has not. In line with this, the scale of analysis can be considered to be of the utmost importance when examining the links between economic development and transport investments. Currently, the literature mainly focuses on the overall regional employment impact or changes in areas of geographical proximity to the investments in question, and in some cases control areas are used to document changes. There is, however, a risk that, by only focusing on parts of the urban area, the general impact of transport investments is concealed. Therefore, it is important to be aware of the extent to which employment growth in, for example, metro-served areas is a result of general employment growth in a region, increasing employment only in metro-served areas, or a result of a wider restructuring of employment on an intra-regional scale. Thus, to understand better the relation between investments in transport and employment growth, the local, regional and intra-regional outcomes all need to be analysed. In the remainder of this paper, the analysis and discussion will address the potential risk of overestimating the economic outcome of investment activities due to the relocation dynamics of intra-urban economic activity from, for example, non-metro-served areas to metroserved areas if only some parts of the relevant urban area are included in the analysis.
spatial challenges, employment restructuring is analysed in two ways. First, we assess the employment restructuring by summarizing and mapping employment growth on various intra-urban scales. Secondly, we analyse employment growth in metro- and non-metro-served areas using OLS regression models. This will provide empirical evidence of intra-urban restructuring of employment by workplace to illustrate the importance of the intra-urban scale. The models do not aim at estimating the direct effect of large infrastructural investments on employment growth within an urban area or seek to establish clear links between these investments and employment changes. Rather they serve to demonstrate how employment growth has developed differently within and outside metro served areas since the metro line in Copenhagen was decided and introduced. The remainder of the paper is structured as follows. In the following section, the links between investments in transport and the labour market are introduced in more detail, and earlier empirical findings are briefly reviewed. In the third section, the present case study of Copenhagen and the Copenhagen Metro and their context of development are presented. The following section introduces the data and the methodological approach chosen for the study. In the fifth section, intra-urban employment growth in Copenhagen between 1992 and 2012 is analysed, while the final section summarises and concludes the paper's findings. 2. Investments in urban transport and their links with the labour market As globalization has developed, it has been increasingly realized that transport is a critical determinant of both the performance of the urban economy and the attractiveness of the city as a place to live and work. Apart from the apparently positive effects of these investments, their wider impacts have gained increasing political and academic attention (Banister and Thurstain-Goodwin, 2011; Knowles and Ferbrache, 2016; Vickerman, 2007). Studies have highlighted that, under certain conditions, urban public transport investments in light rail and underground metro systems can act as catalysts for urban development and redevelopment (Gospodini, 2005). However, despite the development potential, only a few studies have addressed the impacts ex-post. In an extensive review, Gibbons and Machin (2006) stressed that existing research and studies specifically focusing on the impact on the labour market are limited in scope. According to Gibbons and Machin (2006), transport plays a three-way role in relation to labour markets, first, by affecting workers' behaviour and the labour supply; secondly, by affecting firms' decisions and the demand for labour; and thirdly, and as a result of these changes, by changing the equilibrium between supply and demand in the labour market. Two potential labour market impacts are widely acknowledged to occur when transport infrastructure is improved. The first relates to the overall expansion of the labour market catchment area. When transport accessibility is improved, workers are willing to commute longer distances, more people will enter the labour market, and jobs are potentially relocated to more accessible and higher productivity areas (Knowles and Ferbrache, 2016; SACTRA, 1999; Vickerman, 2007). Secondly, improvements in transport infrastructure influence spatial relationships, allowing potential agglomeration effects to occur (Graham, 2007; Venables, 2007; Melo et al., 2013). Agglomeration effects increase with the level of spatial proximity and concentration, and improving the transport network affects the level of concentration and density. This occurs either by increasing the effective density of an area by bringing workers and firms closer together or by relocating firms and/or workers from lower to higher productivity areas (Banister and Berechman, 2001; Wangsness et al., 2016). Regarding the effects on local employment, previous ex-post studies of investments in urban rail-based public transport have shown mixed results. Some studies find evidence of positive employment impacts. In
3. The case of Copenhagen and the Copenhagen Metro The Greater Copenhagen Region is the largest city-region in Denmark. In 2012 it had close to two million inhabitants, while the City of Copenhagen (the central municipalities) had just about 0.7 million inhabitants. Before the opening of the metro, public transport in the region was based on an extensive bus network and “S-bane” commuter train routes connecting the central city to five suburban “Finger Plan”corridors to the south west, west and north west, and two regional commuter train routes from Roskilde in the west and Helsingør in the 22
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
workplace addresses within 600 m walking distance were found (see Fig. 1). Earlier findings have shown that commercial development is most likely to be concentrated within 400 m of a transit station (Banister and Thurstain-Goodwin, 2011; Guerra et al., 2012), but also that the distance will vary from place to place depending on the local context. In this study, a walking distance of 600 m was chosen because urban development in the Greater Copenhagen area is restricted by planning principles near transit stations (Stationsnærhedsprincippet). Although the planning principles place restrictions on development, they allow more dense urban development within 600 m of transit stations. To test the walking distance of the catchment areas, larger catchment areas were also generated, but no significant employment effects were found beyond the 600-m limit. The metro-station catchment areas were divided into five spatially coherent areas based on their primarily urban function, local characteristics and the opening of the metro. The areas are referred to as metro-served areas (MSA), and the division is shown in Fig. 1. “MSA Inner City” is located in the old historical centre and is dominated by a high density of employment. “MSA Christianshavn” is located just south of the harbour and is a more mixed residential and commercial area characterized by large brown-field development sites near the harbour front. “MSA Frederiksberg” is mostly a residential area with a small number of small-scale brown-field sites. “MSA Ørestad” is a green-field development area with mixed land use. “MSA Amager East” is a diverse area dominated by mixed land use in the north, the airport in the south and large residential areas in between. Due to methodological problems in measuring the exact employment changes in the catchment area of the metro station in the airport, the station is not included in the analysis.1 It is though important to state that the airport has grown considerably in the studied period. In the analysis, the employment changes in the metro-served areas are compared to those in the nonmetro-served areas of the city. No specifically matched control areas are therefore used. This is because the metro-served areas cover a large part of the city, meaning that it was not possible to find comparable areas. This is especially the case for MSA Inner City and the green-field development in MSA Ørestad. National and regional trends in development are also included to benchmark the changes in employment. Ordinary Least Squares (OLS) models have been estimated to address intra-urban employment growth in the time period ten years after the metro was opened (2002−12). To test for potential differences in development trends before and after the financial crisis, OLS models have also been estimated for the first five years after the opening (2002–07) and the subsequent five years (2007–12). The main spatial level of analysis in the models is 500 × 500 m grid cells covering the entire City of Copenhagen. Grid cells, and not the metro-served areas described above, were chosen because on a small scale the existing administrative spatial divisions are outdated. Moreover, spatial variation in size in existing divisions makes it difficult to define the metroserved areas properly. Based on matched employee-employer data, all workplaces and numbers of employees have been aggregated for each grid cell. Due to the small number of workers in some peripherally placed cells, it was decided to merge cells with few employees with neighbouring cells. Following the merging exercise, all grid cells consist of at least 150 employees, with the City of Copenhagen being covered by 269 cells. The dependent variable is change in employment, defined as the absolute change in employment between 2002 and 2012. Absolute change has been chosen as the development variable instead of the employment growth rate because cells with few employees had a large influence and therefore caused problems. The main indicator we wish
north. The political decision to build the metro was taken with the Danish government's Ørestad Act of 1992. The investment marks a radical shift in Danish transport and economic development policies and contrasts greatly with policies in earlier decades that were dominated by a narrow focus on favouring car traffic and supporting equal economic development in the entire country. The key idea of the metro was to connect the less accessible southern part of the City of Copenhagen with the inner city and to develop a new large and attractive 310hectare business district, Ørestad, in a greenfield site owned by the city and the state on land reclaimed from the sea in the west of Amager. In 1996 the locations of the line and stations were approved for the first development phases of the metro, and construction started the same year. The metro consists of two lines in western and eastern Amager which are connected though the inner city and the western part of the City of Copenhagen (see Fig. 1). The metro system has 22 stations, of which nine are underground, and it is connected to regional or inter-city trains at four stations. The first phase connecting Ørestad to the inner city opened in 2002, and the second phase connecting the inner city with the western part of the city opened in 2003. The last phase connecting the airport opened in 2007. 4. Research methodology and data There is a contrast between the expected employment impacts of investments in urban public transport and the mixed employment impacts essentially found in ex-post studies. This paper address this issue by exploring the changing geographies of intra-urban employment ten years either side of the opening of the Copenhagen Metro. To assess local development we use employment growth by workplace as a proxy for intra-urban performance. The drawbacks of using employment as the dependent variable are that employment growth does not necessarily mirror the impact on productivity or the transformation in the industrial structure (Eriksson et al., 2017). However, gaining further knowledge of the spatial restructuring of employment growth by workplace is crucial because transport investments are often politically justified on their ability to generate jobs. The aim of the paper is to analyse intra-urban spatial employment structures within and between metro- and non-metro-served areas and to address the scale of the employment restructuring. Therefore, the study does not address the causality of the transport impacts or the direct effect of the metro investment. Employment growth in the period of the study is analysed in two steps. First, employment growth by workplace is addressed descriptively by summarizing them on different spatial scales and by mapping the intra-urban changes on a small scale using kernel densities estimations. Secondly, employment growth in metro- and non-metroserved areas is analysed through regression models. The analysis is based on individual register-based employment data with detailed information on workplace addresses in Copenhagen between 1992 and 2012. All data were obtained from Statistics Denmark, which holds detailed information on workplace addresses for all individuals. The data consist of all employed individuals in the age range of 16–65, and the high quality of the data on the registered workplaces makes it possible to analyse fully all changes in employment by workplace in Copenhagen in this period. The spatial scale is a key issue when analysing employment dynamics. The focus in this paper is on the local and intra-urban development, and the main study area is therefore the City of Copenhagen. This city is normally defined as the municipalities of Copenhagen and Frederiksberg (which is surrounded by Copenhagen), but in this analysis the neighbouring municipality of Tårnby is also included due to the layout of the metro. To address local employment growth, local catchment areas were identified and generated around each metro station, as none of the existing administrative intra-urban divisions of the city was found suitable for the analysis. Based on a detailed road and path network (the Danish Address and Road Database), all
1 During the period of the study, a large proportion of employment moves in and out of the 600-meter catchment area. This is mainly due to changes in the work addresses where people are registered. Also a large proportion of those who work geographically at the airport are registered at a different work address. Approximately 23,000 people are directly and indirectly employed at the airport.
23
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 1. The Copenhagen Metro and metro-served areas in Copenhagen.
impact. However, the controls were not significant, and they were therefore omitted from the final analysis. The goodness of fit tests was generally satisfactory but showed minor issues with heteroscedasticity and spatial autocorrelation in the models. The problem with heteroscedasticity is mainly due to there being few cells where large employment changes are taking place, especially in some metro-served areas, and we therefore decided not to treat them as outliers but to keep them in the models. Running the models with robust standard errors did not change the direction of the impacts but lowered the effect remarkably and therefore we chose to keep the models as OLS models.
to assess is the geographical placement of the cell in relation to the metro stations. Metro-served areas in the models are defined as cells where > 50% of employees' workplaces were located within 600 m of a metro station, and a dummy variable was used as a proxy for the spatial placement. Identifying the metro-served areas in this way causes a degree of uncertainty but was chosen in favour of the existing administrative spatial divisions for pragmatic reasons. To address the different opening years of the metro and the diverse influence in the different parts of the metro corridor, three dummy variables were constructed. The first dummy consists of all metro-served areas, the second of the metro-served areas that opened in 2002/2003 (excluding MSA Amager East), and finally five dummies were constructed for the five different parts of the metro corridor. The three models (A + B + C) are run for each time period using the dummies. All variables in the models are defined in further detail in Appendix A. In explaining changes in employment, we included several control variables in the model. First of all, we included earlier employment growth for 1992–2002, based on the assumption that the opening of the metro has changed the internal patterns of such growth. Secondly, we included the existing employment density in 2002 because the pre-existing distribution of jobs may influence where future jobs are created due to the presence of localization and urbanization economies. Thirdly, we also included a control for human capital defined as the share of employees in 2002 with at least a master's degree. The control is included because a number of studies have shown that a concentration of human capital is important for urban development because it generates spill-over effects and also triggers future concentrations of human capital (see e.g. Hansen and Winther, 2010 on Copenhagen). Initially we included controls for a mix of industries to test whether preexisting concentrations of specific industries such as KIBS had an
5. Intra-urban restructuring and employment growth In the following, employment growth ten years before and after the opening of the metro is analysed, first with a broad focus on the scale of the changes, and secondly with a narrower focus on the differences in growth patterns between and within metro-/non-metro-served areas ten years after the metro was opened. 5.1. Intra-urban employment growth Since the early 1990s, Copenhagen and the largest Danish cities have revitalised themselves around the knowledge and service economy (Hansen and Winther, 2012). This revitalisation is in great contrast to the 1970s and 1980s, which were dominated by a relocation of employment from the largest cities to small and medium-sized cities and from the central city to suburban municipalities in the Greater Copenhagen Area (Illeris, 1997; Winther, 2007; Hansen and Winther, 2010). From the early 1990s the direction of relocations changed, and during the 1990s and 2000s the four largest Danish cities had 24
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Table 1 Employment growth 1992–2012 in metro-served areas compared to changes in City of Copenhagen and Denmark. Employment in absolute numbers
Denmark Aarhus, Odense & Aalborg Greater Copenhagen Area City of Copenhagen - Non-metro-served areas - Metro-served areas - MSA Inner City - MSA Frederiksberg - MSA Christianshavn - MSA Ørestad - MSA Amager East
Change in employment in pct.
1992
1997
2002
2007
2012
92–97
97–02
02–07
07–12
92–02
02–12
2,558,205 329,284 880,548 340,488 248,191 92,297 50,441 22,371 9966 579 8940
2,649,855 346,631 918,723 347,176 257,432 89,744 47,825 22,197 10,137 828 8757
2,727,566 362,442 963,061 369,430 274,624 94,806 46,569 23,226 14,531 1601 8879
2,853,149 391,368 1,024,911 384,839 281,056 103,783 49,022 21,827 16,750 7788 8396
2,658,573 375,143 997,598 391,238 279,751 111,487 47,659 21,683 20,018 14,896 7231
3.58 5.27 4.33 1.96 3.72 −2.77 −5.19 −0.78 1.72 43.01 −2.05
2.93 4.56 4.82 6.41 6.68 5.64 −2.63 4.64 43.35 93.36 1.39
4.6 7.98 6.42 4.17 2.34 9.47 5.27 −6.02 15.27 386.45 −5.44
−6.82 −4.15 −2.69 1.66 −0.46 7.42 −2.78 −0.66 19.51 91.27 −13.88
6.62 10.07 9.37 8.5 10.65 2.72 −7.68 3.82 45.81 176.51 −0.68
−2.53 3.5 3.59 5.9 1.87 17.59 2.34 −6.64 37.76 830.42 −18.56
evident in recent years. Only 1600 people were employed in MSA Ørestad in 2002, but in 2012 the number was close to 15,000. In contrast to the slow start of Ørestad is the development of MSA Christianshavn. In these areas a large increase in employment is seen in the years before the opening of the metro and the last years of the period of the study. The situation here has to be understood in the context of the divergent development policies of the city government (Majoor, 2008) and the local context of the development. These areas are examples of brown-field development benefiting from the redevelopment of the harbour front. Until now, intra-urban employment growth has been described based on the aggregated catchment areas of the metro. To overcome the spatial constraints of using the defined metro-served areas and to address the broader changes in employment, employment densities and changes are mapped on a small scale for the entire city in Figs. 3, 4 and 5 using kernel density estimates based on employment data on 100 × 100 m grid cells. The maps show two important aspects of this intra-urban development. First of all, a clear concentration of employment is evident in the city centre in 2012, with smaller employment concentrations in a few specific areas in the city shown in Fig. 3. Secondly, a large intra-urban variation is seen in employment density change between 1992 and 2012 with both increasing and decreasing density trends (see Figs. 4 and 5). Between 1992 and 2002, a decreasing density is mostly seen in the centre whereas densification is seen in small pockets around the city and especially in some of the redeveloped harbour areas. The development around the metro stations also shows divergent trends. Where densification is seen around few metro stations, especially the metro station Christianshavn in MSA Christianshavn and metro station Frederiksberg in MSA Frederiksberg, a decreasing density is seen around the metro stations in the centre. In Fig. 5, a densification of employment between 2002 and 2012 is, apart from development around the National Hospital and the University Park in the northern part of the city, mostly seen in specific parts of the metro corridor and again in some of the redeveloped areas of the harbour front. This highlights two important aspects of the intra-urban employment development. Firstly, there seems to be a substantial spatial restructuring in the city in the period, and secondly there are large local variations within most of the MSA's. In MSA Christianshavn densification is seen around both metro stations, while no clear pattern and both an increasing and decreasing density is seen around the metro stations in MSA Frederiksberg and MSA Inner City. In MSA Ørestad, employment densification is mainly taking place around two of the metro stations. From the first part of the analysis, it can be seen that employment growth was higher in metro-served areas than in non-metro-served areas in the ten years after the opening. By addressing employment growth in the different metro-served areas, it is also clear that large local differences in employment growth took place and that by mapping them on a micro-scale, a mixed pattern of increasing and decreasing
employment growth above the national average. Whereas 26% of all national employment was located in the four largest Danish cities in 1992, this increased to 29% in 2012. In Table 1 employment growth in 1992–2012 is summarized ranging from the national and urban level down to the metro's catchment areas. The table shows that overall employment growth in the City of Copenhagen differs from trends at the national level and in the other three largest cities in Denmark (Aarhus, Odense and Aalborg) throughout the period. This is also seen in the period 2007–2012, when, despite the financial crisis, the City of Copenhagen had a small growth in employment, while national employment declined markedly. Focusing on intra-urban employment growth in the City of Copenhagen, Table 1 reveals that employment growth in the metroserved areas differs considerably from trends in the rest of the city. In the ten years after the political decision to build the metro was taken and until the opening of its first stage in 2002, employment rose by 2.6% in the metro-served areas, compared to employment growth of 10.7% in the remainder of the city. In the ten years after the metro opened, this picture changed, and the metro-served areas had dramatically greater employment growth (17.6%) compared to the rest of the city (1.9%). With more than three out of four new jobs located in the metro-served areas in this period, the redistribution of employment in Copenhagen has been considerable in this period. From 2002 to 2012 the level of employment in Copenhagen located in metro-served areas increased from 24.9% to 28.1%. Despite overall employment growth in the metro-served areas in 2002–2012, employment growth is unevenly distributed between the different areas. Fig. 2 displays annual growth in employment in the metro-served areas. It can be seen that, while employment grew in the western corridor south of the harbour (MSA Ørestad and MSA Christianshavn), employment levels in the other corridors was constant or decreased slightly throughout this period. The developments observed in MSA Amager East and MSA Frederiksberg were not unexpected because they mostly consist of already existing or newly developed residential areas. The decrease in employment around the metro stations in MSA Inner City is more surprising, and it contrasts with the development patterns seen in several earlier ex-post studies of metro and light rail investments from other European cities. This indicates that the implementation of the metro in Copenhagen do not coincide with boosted employment in the inner city but rather has opened up new economic spaces in the city, where new and redistributed developments are taking place. The levels of employment growth in MSA Ørestad and MSA Christianshavn are examples of this type of development. As already stressed, the development of Ørestad as a new modern business district was one of the main aims of investing in the metro. Despite a slow start in developing Ørestad, which was mostly driven by the relocation of the Danish Broadcasting Company and larger public investments in universities (Knowles, 2012), a significant increase in employment is 25
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 2. Yearly employment growth in 1992–2012 in metro-served areas in City of Copenhagen.
Fig. 3. Employment density in City of Copenhagen 2012. 26
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 4. Employment density change in City of Copenhagen 1992–2002.
growth, indicating that employment growth prior to 2002 influences the employment growth after 2002. Concerning employment growth between 2002 and 07 and 2007–12, we find the same trend in the two periods, but with a higher positive relationship between employment growth and the metroserved areas in the second period. Moreover, important differences are found between models 2C and 3C, where the focus is on different parts of the metro corridor. In the first five years from 2002 to 2007, a significant positive relationship between employment growth and MSA's are found in MSA Ørestad and MSA Inner City while a significant negative relationship is identified in MSA Amager East, where the metro opened later. From 2007 to 2012 a significant positive relationship is still seen in MSA Ørestad and MSA Christianshavn, but not in MSA Inner City. To check for spatial autocorrelation we use a Moran's I test. The test show a small but significant clustering of the residuals in four of the models (model 1A, 1B, 3A and 3C), indicating that the residuals are not randomly distributed. To address the autocorrelation, a Getis-Ord hotspot analysis (Gi* statistic) was done for each set of the residuals. Based on the analysis, a spatial concentration of high values was identified for each model in cells covering Ørestad (for all four models) and the inner harbour (model 3A and 3B). To control for these hot-spot-areas, four different dummies were included in the four models. By doing so, the spatial autocorrelation becomes insignificant, while the results remain robust. In general, the models demonstrate a positive relationship between employment growth and the metro-served areas in the case of Copenhagen. In particular, the metro-served areas in Ørestad and in the central parts of the city seem to have benefitted from employment
employment density is seen. This emphasizes the importance of scale when adressing employment restructuring in metro-served areas. Employment growth is not taking place in all metro-served areas, but primarily in the redeveloped areas south of the harbour and in the new town development in Ørestad. The metro-served areas in the inner city are losing employment relatively, as are some of the other parts of the metro corridor.
5.2. Employment growth in metro- and non-metro-served areas Table 2 presents the output of the regressions of three periods after the metro opened. Concerning employment growth in the first ten years after the opening, we find a positive relationship between geographically being located in a metro-served area and employment growth. The relationship is largest and most significant in models 1B and 1C, indicating that the timing of the opening and spatial variation inside the metro corridor are important for understanding the relationship. Model 1C highlights the earlier descriptive findings and underlines the fact that, while a positive significant relation is seen in the metro-served areas of MSA Ørestad and MSA Christianshavn, a significant negative relation is seen in the metro-served areas of MSA Amager East. Controlling for the level of human capital shows that the pre-existing level has a significant, positive, but limited contribution to employment growth. Controlling for employment density shows a negative but not highly significant relation in some of the models. This indicates that to some extent employment growth is taking place in new and lower density areas of the city. However, this should be seen in relation to the positive, but not greatly significant impact of earlier employment 27
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Fig. 5. Employment density change in City of Copenhagen 2002–2012.
employment structures in Copenhagen differ between areas served by a metro and areas that are not. This is done by using register data along with catchment areas based upon distance to metro stations. Several conclusions can be drawn from the analysis. First and foremost, the analysis shows that scale is of great importance when understanding and questioning the changes within and between metro- and nonmetro-served areas. By addressing the spatial restructuring on an aggregated level, employment growth was significantly higher in metroserved areas compared to non-metro-served areas in Copenhagen in the first ten years after the metro opened. Addressing the employment changes on a small scale also stressed that employment growth was unevenly distributed along the different parts of the metro corridor. Especially in the west of Amager, the introduction of the metro seems to occur simultaneously with employment growth in areas that, before the introduction of the metro, were poorly connected to the rest of the city. Despite significant improvements in transport accessibility in the inner city, the inner city lost employment relative to other areas in this period. The maps in Figs. 4 and 5 provided a visual impression of where increases in employment density have taken place within the City of Copenhagen. Employment density has mainly increased along the metro line and by the harbour front, although a few exceptions can be identified. Just as importantly, by zooming out from the metro line and observing the development across the city, considerable decreases can also be identified. This emphasizes the importance of scale when discussing the effects of urban investments. Secondly, Fig. 2 and Table 1, underlined by the various OLS models, demonstrate that not only is scale an important issue to keep in mind when assessing the impact of transport investments, but so is the time variable. The analysis shows that areas around the metro grew in
growth during the period in question. However, the models do not allow us to assign this positive relationship as an outcome of the introduction of the metro. Most likely, employment growth in metroserved areas is a combination of changing economic structures caters for new location dynamics, investments in the metro and a dedicated political focus on developing and redeveloping specific areas of the city. Also, the models suggest that some metro-served areas of the city, such as Amager East and Frederiksberg in the early period and the inner city in the later period, show negative relationships between employment growth and being located in a metro-served area. This highlights the important question of to what extent the restructuring of employment in the city is caused by a general restructuring of the economy catering for new industries and new location dynamics, the creation of new employment, the emergence of the metro, or by redistribution of already existing employment from one area to another. Lastly, the model also reflects what has already been documented in the case of Copenhagen (Hansen and Winther, 2010), namely that economic development is closely linked to the level of human capital. Here it is shown that, even when looking at small areas, such as metro catchment areas in Copenhagen, the picture is the same: an increase in employment opportunities is more likely to be found in places that already have a high level of human capital. This stresses the difficulties in isolating any direct effect of investments in the metro and emphases that employment trends in the period should rather be seen as a series of effects with different degrees and sorts of influence.
6. Conclusion The aim of this paper has been to examine how intra-urban spatial 28
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Table 2 OLS models on employment growth in the City of Copenhagen 2002–2012. 2002–12 Model 1A Earlier growth 1992–2002 Employment density 2002 Human Capital 2002 All MSA
⁎
1.522 (0.826) −65.534 (53.696) 14.634⁎⁎⁎ (4.968) 227.792⁎ (121.346)
2002–07 Model 1B ⁎⁎
1.790 (0.805) −98.500⁎ (52.696) 11.877⁎⁎ (4.880)
MSA Ørestad MSA Amager East MSA Frederiksberg MSA Inner City MSA Christianshavn
R2 Adj. R2 F-test Mean VIF Moran's I N
1.108 (0.724) −73.005 (48.551) 10.990⁎⁎ (4.528)
Model 2A 0.832 (0.513) −6.152 (33.356) 11.063⁎⁎⁎ (3.086) 52.126 (75.380)
620.614⁎⁎⁎ (142.544)
First MSA
Intercept
Model 1C
2007–12
255.599 (336.405) 0.060 0.046 4.198⁎⁎⁎ 1.12 0.021⁎ 269
473.296 (331.635) 0.111 0.098 8.247⁎⁎⁎ 1.15 0.024⁎⁎ 269
Model 2B ⁎
0.986 (0.506) −24.583 (33.110) 9.765⁎⁎⁎ (3.066)
Model 2C
Model 3A
Model 3B
Model 3C
0.741 (0.475) −30.243 (31.851) 9.338⁎⁎⁎ (2.971)
0.690 (0.568) −59.382 (36.953) 3.571 (3.419) 175.667⁎⁎ (83.509)
0.804 (0.562) −73.917⁎⁎ (36.777) 2.111 (3.406)
0.366 (0.530) −42.763 (35.543) 1.652 (3.315)
275.073⁎⁎⁎ (89.563) 2915.935⁎⁎⁎ (331.343) −466.888⁎⁎⁎ (171.295) −78.745 (183.464) 444.757⁎ (252.300) 642.481⁎⁎⁎ (245.721) 368.130 (305.372) 0.304 0.282 14.171⁎⁎⁎ 1.13 0.003 269
−67.068 (208.975) 0.070 0.056 4.972⁎⁎⁎ 1.12 0.012 269
46.037 (208.372) 0.101 0.087 7.374⁎⁎⁎ 1.15 0.016 269
345.541⁎⁎⁎ (99.482) 1273.538⁎⁎⁎ (217.374) −336.523⁎⁎⁎ (112.376) −105.880 (120.359) 504.979⁎⁎⁎ (165.518) 133.666 (161.202) 118.452 (200.336) 0.232 0.208 9.807⁎⁎⁎ 1.13 0.004 269
322.667 (231.511) 0.028 0.013 1.893⁎ 1.12 0.037⁎⁎⁎ 269
427.259⁎ (231.449) 0.055 0.040 3.825⁎⁎⁎ 1.15 0.034⁎⁎⁎ 269
1642.398⁎⁎⁎ (242.570) −130.364 (125.402) 27.135 (134.310) −60.222 (184.704) 508.816⁎⁎⁎ (179.887) 249.679 (223.557) 0.185 0.160 7.388⁎⁎⁎ 1.13 0.016 269
Standard errors in parentheses. ⁎ p < 0.10. ⁎⁎ p < 0.05. ⁎⁎⁎ p < 0.01.
employment in Ørestad, the initially expected development did not take place there due to a political shift in the period towards a more liberal and market-based planning regime that especially favoured development of the harbour front. Secondly, it is important to understand the local context and the preconditions for the development. Using the OLS models it was shown that employment growth took place especially in areas that already had a high level of human capital. Last but not least, an understanding of the location preferences of firms seems to be important. Although not addressed directly in the present analysis, employment growth is taking place especially within the service and knowledge-intensive sectors, and there are signs of changed location preferences favouring the harbour front and some of the metro-served areas. Therefore, understanding how the location preferences in these sectors are directly affected by investments in transport is an important task for future research.
different time periods, some taking off parallel with the construction of the metro, while others started to blossom at a later stage. This stresses the difficulties in assigning employment changes to an outcome of a large transport investment like the Copenhagen Metro and moreover suggests that results of ex-post studies are highly sensitive to when they are carried out. Thirdly, the interpretations of the outcome of this study support earlier findings showing that a range of necessary conditions need to be in place to stimulate economic development when large public transport investments are made (Banister and Berechman, 2001). In the case of Copenhagen especially three conditions seem to be of great importance when assessing changes to employment. The first relates to the availability of suitable land/space and supporting planning policies. In Copenhagen, employment growth is not random but is to a large degree limited by the existing urban structures of the built environment, taking place especially in brown-field and green-field development areas. Therefore, the development in employment structures needs to be understood in the context of how former industrial areas are becoming available with the transformation towards the service and knowledge economy. Closely related to this, employment changes also need to be seen in the context of supporting planning policies and how these policies change throughout the period. Despite a large increase in
Funding This work was supported by Metroselskabet (108567) as part of the PhD-project “The wider socio-economic and spatial impacts of urban transport investments”.
Appendix A. Variables used in the OLS regressions
Variables
Definition
Dependent variables Employment growth 2002–12 Employment growth 2002–07 Employment growth 2007–2012
Absolute employment change in 2002–12 Absolute employment change in 2002–07 Absolute employment change in 2007–12 29
Journal of Transport Geography 70 (2018) 21–30
K. Bothe et al.
Metro served areasa All MSA First MSA MSA Ørestad MSA Inner City MSA Frederiksberg MSA Amager East MSA Chr/IB Controllers Earlier growth 1992–2002 Employment density 2002 Human Capital 2002
Dummy Dummy Dummy Dummy Dummy Dummy Dummy
variable variable variable variable variable variable variable
indicating indicating indicating indicating indicating indicating indicating
all metro served grid cells grid cells that where metro served in 2002/2003 the metro served grid cells in Ørestad the metro served grid cells in Inner city the metro served grid cells in Frederiksberg the metro served grid cells in Amager East the metro served grid cells in Christianshavn
Employment change (%) in 1992–2002 Total number of people employed in 2002 (log) Share (%) of workers in 2002 with at least a master's degree (ISCED 5A)
a A grid cell is defined as a metro-served grid cell if > 50% of the people employed in the cell have a work address located within 600 m walking distance of a metro station.
Development. Edward Elgar, pp. 259–272. Illeris, S., 1997. The changing location of service activities in the Copenhagen Region. Geografisk Tidsskrift-Danish J. Geogr. 1997, 120–142. Knowles, R.D., 2012. Transit oriented development in Copenhagen, Denmark: from the finger plan to Ørestad. J. Transp. Geogr. 22, 251–261. Knowles, R.D., Ferbrache, F., 2016. Evaluation of wider economic impacts of light rail investment on cities. J. Transp. Geogr. 54, 430–439. Majoor, S., 2008. Progressive planning ideals in a neo-liberal context, the case of Ørestad Copenhagen. Int. Plan. Stud. 13, 101–117. Mayer, T., Trevien, C., 2015. In: Papers, C.E.P.R.D. (Ed.), The Impacts of Urban Public Transportation: Evidence from the Paris Region, pp. 31 INSEE Working paper G2015/0. Meijers, E., Hoekstra, J., Leijten, M., Louw, E., Spaans, M., 2012. Connecting the periphery: distributive effects of new infrastructure. J. Transp. Geogr. 22, 187–198. Mejia-Dorantes, L., Lucas, K., 2014. Public transport investment and local regeneration: a comparison of London's Jubilee Line Extension and the Madrid Metrosur. Transp. Policy 35, 241–252. Melo, P.C., Graham, D.J., Brage-Ardao, R., 2013. The productivity of transport infrastructure investment: a meta-analysis of empirical evidence. Reg. Sci. Urban Econ. 43, 695–706. Nelson, A.C., Sanchez, T.L., 1997. The Influence of MARTA on Population and Employment Location. Paper Presented at the 76th Annual Meeting of the Transportation Research Board, Washington, DC. Padeiro, M., 2013. Transport infrastructures and employment growth in the Paris metropolitan margins. J. Transp. Geogr. 31, 44–53. SACTRA, 1999. Transport and the Economy - Final Report. Standing Advisory Committee on Trunk Road Appraisal, London. Scott, A.J., 2008. The Social Economy of the Metropolis. Oxford University Press, Oxford. Venables, A.J., 2007. Evaluating urban transport improvements: cost-benefit analysis in the presence of agglomeration and income taxation. J. Transp. Econ. Pol. 41, 173–188. Vickerman, R., 2007. Recent evolution of research into the wider economic benefits of transport infrastructure investments. In: JTRC Discussion Paper. 2007. Vickerman, R., 2008. Transit investment and economic development. Res. Transp. Econ. 23, 107–115. Wangsness, P.B., Rødseth, K.L., Hansen, W., 2016. A review of guidelines for including wider economic impacts in transport appraisal. Transp. Rev. 1–22. Winther, L., 2007. Location dynamics of business services in the urban landscape of Copenhagen: imaginary spaces of location. Belgeo 2007, 51–72.
References Banister, D., Berechman, Y., 2001. Transport investment and the promotion of economic growth. J. Transp. Geogr. 9, 209–218. Banister, D., Thurstain-Goodwin, M., 2011. Quantification of the non-transport benefits resulting from rail investment. J. Transp. Geogr. 19, 212–223. Beyazit, E., 2015. Are wider economic impacts of transport infrastructures always beneficial? Impacts of the Istanbul Metro on the generation of spatio-economic inequalities. J. Transp. Geogr. 45, 12–23. Bollinger, C.R., Ihlanfeldt, K.R., 1997. The impact of rapid rail transit on economic development: the case of Atlanta's MARTA. J. Urban Econ. 42, 179–204. Cervero, R., Landis, J., 1997. Twenty years of the Bay Area rapid transit system: land use and development impacts. Transp. Res. A 31, 309–333. Docherty, I., MacKinnon, D., 2013. Transport and economic development. In: Rodrigue, J.P., Notteboom, T., Shaw, J. (Eds.), The Sage Handbook of Transport Studies. Sage, London, UK, pp. 226–240. Eriksson, R.H., Hansen, H.K., Winther, L., 2017. Employment growth and regional development: industrial change and contextual differences between Denmark and Sweden. Eur. Plan. Stud. 25, 1756–1778. Gibbons, S. & Machin, S. 2006. Transport and Labour Market Linkages: Empirical evidence, implications for policy and scope for further research. Background paper for the Eddington report to the Department of Transport. Gospodini, A., 2005. Urban development, redevelopment and regeneration encouraged by transport infrastructure projects: the case study of 12 European cities. Eur. Plan. Stud. 13, 1083–1111. Graham, D.J., 2007. Agglomeration, productivity and transport investment. J. Transp. Econ. Pol. 41, 317–343. Graham, S., Marvin, S., 2001. Splintering Urbanism: Networked Infrastructures, Technological Mobilities and the Urban Condition. Routledge, London. Green, R.D., James, D.M., 1993. Rail Transit Station Area Development: Small Area Modeling in Washington, D.C. M.E. Sharpe, Armonk, NY. Guerra, E., Cervero, R., Tischler, D., 2012. Half-mile circle. Transp. Res. Rec. 2276, 101–109. Hansen, H.K., Winther, L., 2010. The spatial division of talent in city regions: location dynamics of business services in Copenhagen. Tijdschr. Econ. Soc. Geogr. 101, 55–72. Hansen, H.K., Winther, L., 2012. The Urban Turn: Cities, Talent and Knowledge in Denmark. Aarhus Universitetsforlag, Aarhus. Holvad, T., Leleur, S., 2015. Transport projects and wider economic impacts. In: Hickman, R., Givoni, M., Bonilla, D., Banister, D. (Eds.), Handbook on Transport and
30