At least since the emergence of the modern welfare state model, targeting local problems through spatially focused interventions became an essential part of political agendas in the developed world (van Gent, Musterd & Ostendorf 2009). More recently, interest in urban policies has also been on the rise in other parts of the world, including in particular the post-socialist countries of Central and Eastern Europe. In Poland, for instance, numerous regeneration projects flourished following accession to the European Union and an initiative promoting a national urban agenda has been launched, coupled with the enacting of the first urban regeneration law.
Set against this background, the primary objective of this paper is to provide a better understanding of urban policy initiatives using evidence drawn from a context where such initiatives are already well-established practice. We have taken a specific look at the distribution of urban policy funds in Germany and tried to find linkages with so-called geographies of disadvantage (Pawson & Herath 2015). In so doing, we identified variables that can be used to characterise territorial structural weakness, and then employed the technique of geographically weighted regression to explore their linkages with the spatial distribution of funds from the urban development support scheme (Städtebauförderung). The paper starts with a brief literature overview, followed by a materials & methods section. The next chapter presents the main empirical findings, and the final section presents our conclusions.
Urban policy funds: targeting disadvantaged places
Despite a recent upsurge in research on cities, there is little consensus in the literature regarding the understanding of the term “urban policy”. In general terms, two largely divergent views could be found at opposite ends of the spectrum. According to one view, the scope of urban policy should be drawn broadly to include virtually all the interventions in a certain way considered as relevant for cities (Glaeser 2012: 111). In other words, transport policy, housing policy as well as social programmes or land use regulation could all be counted, as they are all, to a certain extent, of the utmost importance for urban development. A narrower perspective is to consider urban policy as an intervention specifically and explicitly focused on urban areas, that is, an area-based policy (Cochrane 2007: 3). To avoid confounding these two overlapping yet distinctive concepts, we suggest using the term “urban policy” in the broader meaning, and the term “urban programme” in the more specific one.
The origins of urban programmes in developed and highly urbanised countries are to be sought in the early post-war period. Depending on local contexts, and the aims and orientation of such programmes they were given different names, like urban regeneration, revitalization or renewal. The main rationale behind these various initiatives has been to initiate or catalyse changes in the urban environment through spatially and temporally coordinated actions. Despite having undergone numerous changes over the course of years, urban programmes continue to form an essential part of the political agenda in a number of countries in Western Europe and beyond (de Dekker et al. 2003). In some countries, however, including particularly the United States, where urban issues topped the political agenda in the years following President Johnson’s “War on poverty”, the importance of such initiatives has been on the decline in recent decades (Silver 2010).
The delimitation of “spatially disadvantaged” areas to be targeted by policy interventions has always tended to be a somewhat puzzling task. This is also reflected by the terminology, with different authors referring to locational disadvantages, geographies of disadvantage, disadvantaged places (Darcy 2007; Pawson & Herath 2015), or even “unliveable” and “most deprived” urban areas (van Gent et al. 2009). Neither is there some common understanding on the criteria to be applied in delineating such areas. According to W. van Gent et al. (2009: 55), deprived areas are places where place-based liveability issues (like vandalism, anti-social behaviour, crime etc.) are coupled with, and are assumed to be a source of, sustained economic deprivation. Multiple socio-economic variables and composite scores are used more frequently and are believed to be more appropriate measures of spatial disadvantage than simple income indicators (Pawson & Herath 2015). Examples of such composite measures include periodically revised and updated indices of local deprivation in England (Lee 1999; Rae 2012), combining characteristics related to health, income, employment, education or housing.
The linkages between local needs and the distribution of policy funds have been widely discussed in the literature (e.g. Brennan, Rhodes & Tyler 1999; Lawless 2006; Tunstall & Lupton, 2003). In the UK the subject has received much attention and has also raised a certain degree of controversy particularly with regard to the effectiveness of area-based policies in targeting people in need. However, A. Brennan et al. (1999: 2082) are supportive of such interventions, concluding that “the SRB [Single Regeneration Budget] Fund has been highly responsive to the graduation of relative needs across the LA districts of England”. R. Tunstall & R. Lupton (2003: 26) share this view, arguing that “area targeting […] is a more complete way of reaching the poor than has been claimed by opponents of area-based targeting in the past”.
As for other countries, the evidence on the linkages between the distribution of policy funds and geographies of disadvantage is much more limited. In Germany, the leading nationwide urban initiative since the 1970s has been the urban development support scheme (Städtebauförderung). Resembling the federal structure of the German state, the policy is being jointly implemented by the federation (Bund), the sixteen federal states (Länder) and the municipalities (Gemeinden). According to the German constitution, the responsibility for fostering local development lies primarily with the federal states. However, an exception is made for the urban development support scheme, allowing the federal authorities to intervene with the purpose of promoting equal living conditions (gleichwertige Lebensverhältnisse) throughout the national territory (Walter 2001). The urban development support scheme has traditionally been focused on changes in the physical urban fabric, but more recently socio-economic regeneration has gained in importance. As of 2016, the scheme was composed of five separate but mutually interconnected programmes (Fig. 1), including: “Preservation of urban monuments” (Städtebaulicher Denkmalschutz), „The Social City“ (Soziale Stadt), „Urban restructuring“ (Stadtumbau), „Active city and district centres“ (Aktive Stadt- und Ortsteilzentren), and „Smaller towns and rural municipalities“ (Kleinere Städte und Gemeinden).
A peculiarity of the German allocation system is derived from the federal government structure. In more centralised contexts, like the British one, disadvantaged areas are broken up into fine-grained areas like wards. In Germany, funds are first split among the different federal states, which then provide municipalities with funding for their area-based projects on a competitive basis. The choice of areas designated for participation in urban programmes lies therefore primarily with the municipalities.
Materials and methods
This paper employs the technique of geographically weighted regression (GWR), which was first proposed by C. Brunsdon, S. Fotheringham and M. Charlton back in the 1990s (Brundson et al. 1996). Since then, it has been increasingly used in a variety of spatial applications, including socio-economic geography (e.g. Fotheringham, Charlton & Brundson 2001; Huang, Wu & Barry 2010; Wheeler, Rigby & Huriwai 2006). A distinguishing feature of the GWR is the assumption that regression coefficients may vary locally, rather than being global estimates for the whole dataset. Therefore, it provides a means to capture a spatially differentiated relationship between the dependent variable and the explanatory variables. In many geographical contexts this may better express the actual relationships than standard regression models. Previous studies have demonstrated that spatially disadvantaged areas show a strong tendency towards spatial autocorrelation (Rae, 2012). We used the R package spgwr (Bivand & Yu 2017) to estimate the regressions.
The spatial phenomenon we are interested in is the allocation of funds from the German urban development support scheme (Städtebauförderung). This is measured by means of three variables, including: (1) total funding for the urban development support scheme, (2) funding for the specific programme „Urban restructuring“, (3) funding for the specific programme „The social city“. We include funding for two specific programmes beside the total allocation to capture specific programme-related effects. These two particular programmes were chosen mainly because they have now been in operation for a long period. Data were drawn from the database “Städtebauförderungsdatenbank des BBSR” and made available by the Federal Institute for Research on Building, Urban Affairs and Spatial Development (Bundesinstitut für Bau-, Stadt- und Raumforschung). Each variable includes a sum of per capita funding for the 2006 to 2015 period.
In choosing the explanatory variables, we sought a compromise between theory and operationalisation on the one hand and data availability on the other. In particular, we have had to take into account the fact that certain variables are only available for short time windows, which limits their applicability for the purpose of this analysis. The principle objective of the urban development support scheme, which is derived from the constitution of the Federal Republic of Germany, is to target “structurally weak” areas. However, neither the constitution itself, nor any other German legal act, provides an operational definition of such areas. We therefore referred to a delineation of structurally weak areas prepared by the Federal Institute for Research on Building, Urban Affairs and Spatial Development (Göddecke-Stellman & Wagener 2009). From this source we adopted, with slight modifications, a set of variables measuring territorial structural weakness, or strength. These include in particular: population change (year-to-year), change in employment (year-to-year change in the number of employed persons with social insurance), unemployment rate, municipal tax revenue (total revenue of municipalities from local taxes) and disposable household income (disposable income net of taxes, including social transfers). In addition, we also included the population and municipal debt (Tab. 1). The explanatory variables are averages for either 2006 to 2015 or a slightly shorter period, depending on availability. Data are drawn from the INKAR database (INKAR website …), or from the database of the Federal Statistical Office (Statistisches Bundesamt website …).
Summary information on variables used for the analysis
Source: own calculations based on the INKAR database (INKAR website…) and the database of the Federal Statistical Office (Statistisches Bundesamt website…)
|Min||1st quartile||Median||3rd quartile||Max|
|Urban policy funds per capita (euro)||0.00||29.50||47.40||85.13||370.20|
|Population change (%)||-1.32||-0.40||-0.06||0.19||1.39|
|Unemployment rate (%)||2.10||4.90||7.00||10.08||18.70|
|Employment change (%)||-0.48||1.20||1.73||2.22||5.62|
|Disposable household income (1,000 euro)||1.25||1.45||1.58||1.71||2.95|
|Municipal tax revenue per capita (1,000 euro)||0.29||0.50||0.60||0.71||1.84|
|Municipal debt per capita (1,000 euro)||0.00||0.75||1.26||2.25||17.69|
|Population (1,000 persons)||34.30||105.55||149.10||239.83||3426.80|
Our basic observational unit is the German administrative district called a Kreis, or kreisfreie Stadt (literally “a district-free borough”) in the case of larger towns. In Eurostat’s Nomenclature of Territorial Units for Statistics these districts correspond to the 3rd level of subdivision (NUTS-3). There are in total 402 such units in the database, including 299 districts and 103 district-free cities. The choice of the spatial unit of analysis was mainly driven by data availability.
Results and discussion
We began by estimating a standard ordinary least squares regression (OLS) which gives us overall estimates for the whole dataset (Tab. 2). The OLS explains 57% of variance in the dependent variable, suggesting that a substantial share of explanatory factors is being accounted for in the model. The results show that a higher allocation of urban policy funds was to be found in units with a higher unemployment rate, lower employment and population change, lower population number and lower municipal debt per capita. With the exception of disposable household income and municipal tax revenues, which turned out to be insignificant, these results largely confirm our expectations. High unemployment as well as low population and employment dynamics are features commonly attributed to spatially disadvantaged areas. The negative coefficient of population means that on average small and medium-sized towns were allocated more funds from the urban development support scheme than large cities. In other words, the distribution is a degressive one, but only to a marginal extent, as increasing the population by 100,000 persons decreases funding by just 3 euro per capita. A negative relationship with municipal debt suggests that more indebted municipalities received less funding. This outcome could be likely traced back to a general policy rule, which requires municipalities to provide a one third contribution to project costs, with the remaining two thirds being provided in equal shares by the federation and federal states. For municipalities troubled by high levels of debt, it might turn out to be difficult or even impossible to provide the required contribution. They might therefore be unable to participate in the urban development support scheme, or might be forced to limit the scope of their participation.
Allocation of urban policy funds – OLS regression results
Source: own calculations based on the INKAR database (INKAR website…), the database of the Federal Statistical Office (Statistisches Bundesamt website…), and data provided by the Federal Institute for Research on Building, Urban Affairs and Spatial Development Note: the dependent variable is the total allocation of urban policy funds per capita
|Estimate||Std. Error||t value||Pr(>|t|)|
In the next step, for the same variable, we estimated a geographically weighted regression. Visualising the results, we find that a number of characteristic features of the spatial distribution are well captured by the model (Fig. 2). Significantly higher values in Eastern Germany, but with a lower allocation in and around Berlin are correctly predicted, as are the peaks (“islands”) in some county-free cities in the rest of the country. The model also shows a north/south division of Bavaria, with the southern part recording some of the lowest amounts of urban policy funds nationwide. Somewhat misleadingly, the GWR regression identifies a hot spot in the Ruhr area, which does not stand out clearly in the data. There appear to be some clusters of above-average residuals in Eastern Germany, particularly in Saxony and Thuringia, which means that these areas received more funding than the model would actually suggest. In general, however, the distribution of residuals does not suggest a systematic distortion pattern.
An interesting feature of the GWR is that it provides local estimates of the R2 coefficient. In that way, it allows one to compare the goodness of fit between locations instead of just having an overall estimate. In our case, we find a large variation in local R2 values ranging from 0.25 to 0.78, with a median value of 0.60. There is a clear spatial pattern in local R2 values with a declining gradient from the northeast to the southwest. This also means that the model shows the best fit in areas where most funds were allocated. As for the areas with lower per capita funding, the data appear to be too aggregated to capture the actual hot spots targeted by the policy.
The GWR also enables the estimation of local coefficient values instead of a single overall estimate. As for the intercept, the highest values are to be found in Eastern Germany, from where they decrease gradually, somewhat less to the west than to the northwest and the southwest (Fig. 3). This difference reflects to a large extent the fact that allocation levels tend to be much higher in most East German districts than throughout the rest of Germany. As a general rule, federal funds are split between the federal states according to an algorithm based 70% on population, and 30% on a set of socio-economic variables. For some programmes, however, including in particular the urban restructuring programme and the preservation of urban heritage programme, the funds are first divided between the Eastern and the Western part of Germany on the basis of a predefined proportion. This seems to be the primary reason why districts in Eastern Germany continue to receive an amount of urban policy funds which is higher than one would actually expect taking into account their current socio-economic situation.
Interestingly, the coefficients of all the variables included in the model show a relatively large variability in space. While the overall contribution of population change was negative, we observe the lowest coefficient values (i.e. highest coefficient values in absolute terms) in parts of both Eastern Germany (Saxony) and Western Germany (Lower Saxony/North Rhine-Westphalia). These two areas actually differ quite considerably, with the first being characterised by population decline and a high allocation of urban policy funds, in contrast to the latter.
Coefficient values of the unemployment rate look much like a negative to those of population change, with the highest values to be found in the north and the south. These two areas are, again, quite at odds in terms of both the explained and the explanatory variable. They are also, however, somewhat heterogeneous. The northern area comprises parts of Mecklenburg-Vorpommern with a very high allocation of urban policy funds and high unemployment, as well as a segment of Schleswig-Holstein, where the respective values are lower. The southern area comprises mostly Bavaria, with very low unemployment/allocation in the south and somewhat higher values to the north.
In the case of employment change, the lowest coefficient values (i.e. highest values in absolute terms) are clustering in the northeast, and decrease gradually towards the southwest. The part of Germany with the smallest increases in employment was Mecklenburg-Vorpommern, that is, the northernmost of the eastern federal states. In all the other parts of the country the increases were higher, with southern Bavaria and Lower Saxony being two hot spots of employment gains.
The coefficients of household income show an almost perfect longitudinal gradient. This variable, just to recall, did not turn out to be significant in the OLS model we estimated previously. Neither did the tax revenue, in the case of which the local coefficients create a form of concentric rings.
In the case of municipal debt per capita, the highest absolute coefficient values were to be found in the southeast, from where they decreased towards the northwest. Comparing this distribution with the distribution of observed values, we find that it follows a more or less reversed pattern. That is, the highest coefficient values are found in areas where municipal debt per capita was low, and vice versa.
The last variable, the population number, shows the highest absolute coefficient values clustered in central-western part of Germany. It is a densely populated area with a large number of district-free cities, which in comparison to neighbouring districts tend to have a lower sized population, but a higher allocation of urban policy funds.
In the last part of the empirical section we looked at the distribution of funds from two specific programmes implemented under the framework of the urban development support scheme. The first one, the “Social city” programme, initiated in 1999, is an initiative particularly directed at fostering social cohesion in neighbourhoods with a high risk of exclusion and deprivation. Since the very beginning, there was a single amount of funding to be divided among all the federal states, without any preference given to Eastern Germany. The “Urban restructuring” programme, started in 2002 in Eastern Germany and two years later extended to the western part of the country, has been specifically tailored to address the issue of urban shrinkage. Unlike the “Social city” programme, the funds are first divided between Western and Eastern Germany according to a fixed proportion, before they get allocated to particular federal states.
As in the case of total allocation, we estimated the first OLS models for both programmes (Tab. 3). The results for the “Social city” programme are to some extent surprising. While there was an expected positive relationship with unemployment, we also found unexpected positive linkages with population change and tax revenue. Taking into account also the negative relationship with population number, we conclude that the results largely reflect the fact that the allocation of funds from this programme tended to concentrate in district-free cities (Fig. 4). However, as the model only explained 14% of variance, we must conclude that the spatial distribution has been only partly accounted for in the model specification. There could be, generally speaking, two reasons for this.
Allocation of funds in the in “The social city” programme and the Urban restructuring programme – OLS results
Source: own calculations based on the INKAR database (INKAR website …), the database of the Federal Statistical Office (Statistisches Bundesamt website …), and data provided by the Federal Institute for Research on Building, Urban Affairs and Spatial Development
|“The social city” programme||Urban restructuring programme|
|(Intercept)||1.404 (9.957)||4.355 (20.639)|
|Population change||6.224 (1.835)***||-10.135 (3.804)**|
|Unemployment rate||1.501 (0.311)***||5.547 (0.644)***|
|Employment change||-0.968 (1.112)||-6.790 (2.304)**|
|Household income||-5.259 (5.578)||2.807 (11.563)|
|Tax revenue||16.083 (5.328)**||-13.411 (11.043)|
|Municipal debt||0.242 (0.414)||-2.713 (0.858)**|
|Population||-0.011 (0.003)***||-0.015 (0.007)*|
The first hypothetical explanation would be that that the “Social city” programme does not follow the principle of targeting structurally weak areas. An alternative, and in our view more plausible answer is that that the scale of analysis would have to be more fine-grained to capture the structurally weak urban neighbourhoods actually targeted by this policy.
As for the urban restructuring programme, the results look much more similar to the outcomes from the model for total allocation. We found expected negative coefficients for population change and employment change, as well as a positive one for unemployment. There was also a negative linkage with municipal debt and a weak negative association with population. Almost half of the variance (45%) was explained in the case of this programme, somewhat less than for the total allocation of urban policy funds, but significantly more than for the “Social city” programme.
Finally, looking at the GWR results, we find that the spatial pattern of allocation is quite well reflected for the urban restructuring programme. The spatial distribution of funds earmarked for this programme resembles the total allocation of urban policy funds to a large extent. As one could expect, there is a large cluster of high values in Eastern Germany, while in the western part the situation is regionally differentiated. Again, we conclude that the explanatory power of the model is the strongest in the areas with the highest values of the dependent variable. Local R2 values start from 0.27 at the western boundaries to peak at 0.68 in the east.
As for the “Social city” programme, the model seems to be best fitted to the situation of Eastern Germany, and in particular to its northern part. Local R2 values range from 0.15 in the south (Bavaria), and increase radially towards the west, north and east to reach a maximum of 0.51. For parts of Western Germany, the model actually seems to underestimate the amount of funding allocated. However, the “islands” of significantly higher allocation in district-free cities are for the most part predicted properly. The spatial distribution of funds from the “Social city” programme follows a significantly different pattern from that of the total allocation of funds as well as the allocation of funds from the “Urban restructuring” programme, without large clusters of either high or low values and a more dispersed distribution.
For the past few decades, the use of area-based policies has become one of the ways of addressing problems of disadvantaged urban neighbourhoods. In this paper we looked at the allocation of urban policy funds taking the German urban development support scheme as a case study. Our results show that higher levels of funding were more likely to be found in areas with a high unemployment rate as well as low increases, or losses, in population and jobs. This suggests there was a certain degree of correspondence between the distribution of structurally weak areas and the allocation of urban policy funds. We also found a negative association with municipal debt per capita, leading us to the conclusion that highly indebted municipalities could have difficulties in getting access to funding. Using the technique of geographically weighted regression, we observed that the model shows the best fit in areas where the allocation of funds was the highest, i.e. in Eastern Germany. In parts of Western Germany, on the other hand, the mode explained a substantially lower, but still fairly reasonable, part of the variance. The GWR has also shown that the coefficient values of explanatory variables were significantly differentiated in space. Finally, comparing two specific programmes, we found that the results for the “Urban restructuring” programme largely correspond with results for the urban development support scheme as a whole. Despite changes this programme has recently undergone (Radzimski 2016), it remains focused on structurally weak areas, and more specifically on places affected by long-term demographic shrinkage. In the case of “The social city” programme, on the contrary, only a small portion of variance was explained, and a more fine-grained analysis could help to capture the small-scale areas targeted by this policy.
To conclude, the findings suggest that the German urban development support scheme focused on areas that reveal some characteristics of structural weakness. This is generally in line with the tradition of area-based interventions in Germany, which tended to put more emphasis on cohesion than on competitiveness (Frank 2008). Thus, we could confirm previous observations that the German planning system is only marginally influenced by Anglo-Saxon neoliberalisation trends (Waterhout, Othengrafen & Sykes 2013). Whether this also continues to be the case in the long term, or whether a certain convergence is about to take place, is likely to depend in particular upon the stability of institutional planning arrangements in Germany as well as upon the extent to which the constitutional rule of “equal living conditions” maintains its paradigmatic status in the formulation of spatial strategies and policies.
This research has been financially supported by the National Science Centre of Poland (grant no.: DEC-2013/09/D/HS4/00575).
Bivand R. & Yu D. (2017) spgwr: Geographically Weighted Regression. R package version 0.6-31. Available from: https://CRAN.R-project.org/package=spgwr [accessed: 15.12.2017].
Brennan A. Rhodes P. & Tyler P. (1999) The Distribution of SRB Challenge Fund Expenditure in Relation to Local-area Need in England Urban Studies 36(12) 2069–2084.
Brundson C. Fotheringham S. & Charlton M. (1996) Geographically Weighted Regression: a Method for Exploring Spatial Nonstationarity Geographical Analysis 28(4) 281–298.
Cochrane A. (2007) Understanding urban policy: a critical approach Blackwell Publishing Malden.
Darcy M. (2007) Place and disadvantage: the need for reflexive epistemology in spatial social science Urban Policy and Research 25(3) 347–361.
de Dekker P. Vranken J. Beaumont J. & van Nieuwenhuyze I. V. eds. (2003) On the origins of urban development programmes in nine European countries Garant Publishers Antwerp.
Fotheringham S. Charlton M. & Brundson C. (2001) Spatial Variations in School Performance: a Local Analysis Using Geographically Weighted Regression Geographical & Environmental Modelling 5(1) 43–66.
Frank S. (2008) Stadtentwicklung durch die EU: Europäische Stadtpolitik und URBAN-Ansatz im Spannungsfeld von Lissabon-Strategie und Leipzig Charta Raumforschung und Raumordnung 2 107–177 [in German].
van Gent W. P. C. Musterd S. & Ostendorf W. (2009) Disentangling neighbourhood problems: area-based interventions in Western European cities Urban Research & Practice 2(1) 53–67.
Göddecke-Stellmann J. & Wagener T. (2009) Städtebauförderung – Investitionen für die Zukunft der Städte Informationen zur Raumentwicklung 3/4 181–192 [in German].
Huang B. Wu B. & Barry M. (2010) Geographically and temporally weighted regression for modeling spatio-temporal variation in house prices International Journal of Geographical Information Science 24(3) 383–401.
Lawless P. (2006) Area-based Urban Interventions: Rationale and Outcomes: The New Deal for Communities Programme in England Urban Studies 43(11) 1991–2011.
Lee P. (1999) Where are the socially excluded? Continuing debates in the identification of poor neighbourhoods Regional Studies 33(5) 483–486.
Radzimski A. (2016). Changing policy responses to shrinkage: The case of dealing with housing vacancies in Eastern Germany Cities 50 197–205.
Tunstall R. & Lupton R. (2003) Is Targeting Deprived Areas an Effective Means to Reach Poor People? An assessment of one rationale for area-based funding programmes (CASEpaper 70) London School of Economics London.
Walter K. (2001) Das Bund-Länder-Städtebauförderung – ein Erfolgsmodell vertikaler Politikverflechtung Informationen zur Raumentwicklung 9/10 517–525 [in German].
Waterhout B. Othengrafen F. & Sykes O. (2013) Neo-liberalization Processes and Spatial Planning in FranceGermanyand the Netherlands: An Exploration Planning Practice and Research 28(1) 141–159.
Wheeler B. W. Rigby J. E. & Huriwai T. (2006) Pokies and poverty: problem gambling risk factor geography in New Zealand Health & Place 12 (1) 86–96.
INKAR website (Indikatoren und Karten zur Raum- und Stadtentwicklung): http://www.inkar.de [accessed: 20.09.2017] [in German].
Statistisches Bundesamt website (Die Regionaldatenbank Deutschland): https://www.regionalstatistik.de/genesis/online/log-on [accessed: 15.09.2017] [in German].