The results show that the majority of German city regions have either seen higher increases of residents on benefits in suburban communities (points above the diagonal line in the upper right quadrant) or a relatively stable proportion in the commuter sheds compared to decreasing rates in the core city in the observation period (upper left quadrant). This finding backs the interpretation that residents on benefits are increasingly to be found in the suburbs of these city regions. However, it does not constitute proof of displacement and relocation from inner cities. As such households are provided with housing wherever they apply for benefits, they rarely move across municipal borders once they are granted social housing and benefits. Decreasing rates more often mean that households no longer qualify for benefits due to the successful labour market re-integration of previously unemployed persons (Brenke 2018).

This observation is further substantiated by an analysis of employment figures in the same city regions. Labour market reforms in Germany and the economic prosperity of the 2010s led to an increase of jobs in German city regions, with overall unemployment decreasing. However, social policy analysts frequently point out that a large number of the newly created jobs were in the low-income bracket and thus precarious (e.g., agency work not covered by collective agreements, fixed-term contracts, seasonal work), with the result that ‘working poor’ numbers increased. Such criticism puts a major question mark over the apparent successes of Germany’s labour market reforms (Wimbauer/Motakef 2021). Grabka and Schröder (2019) observe that the share of working poor had already increased a few years earlier and has remained constant since 2008. With a share of 23 %, the proportion of employees in the low-wage sector is very high (Grabka/Schröder 2019: 249). This shift in income levels is further backed by observations of increasing socio-economic disparities, with the lower-middle class increasingly exposed to rising living costs not compensated by higher wages.

The emerging ‘blind spot’ in the available data for this social group highlights the need to go one step further, looking for a method that includes the ‘working poor’ in the equation.

The method found to achieve this aimed to map low-income household concentrations in selected German city regions (‘macro-analysis’) with the goal of equipping interviewers in the project team with initial results for discussion with local experts. Explorative data analysis used several possible indicators available for the whole of Germany from official datasets and research institutions and led to the choice of a 1x1 km grid dataset variable from the research data centre at the RWI – Leibniz Institute for Economic Research. This institute specializes in the publication of socio-economic variables, including aggregated micro-geographic data from private data vendors.

The aggregation procedure summarizes street address data to the 1x1 km grid complying with the General Data Protection Regulation (i.e., counts less than 5 are blacked out).

From the investigated variables (and their potential combinations), we ended up choosing the concept of household purchasing power as the most suitable proxy for relative poverty in comparison to the local and city-regional averages. We prefer this procedure over the usage of a fixed income threshold value since income levels are highly diverse across German city regions. In this context, purchasing power is a most conclusive proxy for the affordability of housing where local deviations (‘cold spots’) hint at relative concentrations of low-income and poorer households. It is defined as net household income from labour, interest payments, rents and leases. It is further adjusted for social transfers. Expenses such as taxes or social security contributions are deducted. Social benefits such as unemployment benefit, family allowances and pensions are added where applicable.

https://fdz.rwi-essen.de/doi-detail/id-107807micromkaufkraftv7.html (18.08.2023).

It is important to note that this data source is based on aggregated street-level data modelled by the market research institute Michael Bauer Research GmbH in cooperation with the data vendor microm. Coming from the Federal Statistics Office, their input data represents purchasing power at municipality level. This is then disaggregated to street level using explanatory variables for local differentiation (e.g., typology, age, status, car variables). The results are validated by a sampling strategy that allows a comparison with the socio-economic survey panel (SOEP) and other survey information. RWI researchers check the validity of the data and their application for socio-economic spatial analysis (Budde/Eilers 2014; Breidenbach/Eilers 2018).

Based on the selection of purchasing power as an easily communicated indicator, the macro-analysis identifies spatial concentrations of low-income and poor households in German urban regions. The analysis can be performed by spatial geostatistical analysis packages that use the Getis-Ord Gi* algorithm, aka ‘hotspot analysis’. It is available in commercial software packages like ArcGIS Desktop or ArcGIS Pro or open-source alternatives like GeoDa or the R statistical software package. In the macro-analysis, the Getis-Ord Gi* method compares the value of a cell with the average of all cells of a city region, moderated by a distance decay function. This means that cells neighbouring the purchasing power cell in question have a greater influence on the weighting algorithm than cells farther away. The Getis-Ord Gi* algorithm uses parameters to specify neighbourhood size (‘zone of indifference’) as well as a linear function for distance decay. We used a neighbourhood size of two kilometres, i.e., two cells, and Euclidian distance to weight the influence of neighbouring cells on the result of a selected cell (Getis/Ord 2010).

The output of this ‘hotspot’ analysis also includes so-called cold spots, places where the local purchasing power deviates negatively from the surrounding area and the city-regional average. Figure 2 shows an example for the city region of Augsburg where red and blue cells are colour-shaded based on the level of significance (99 %, 95 % and 90 % significance). High values express a high statistical robustness of the deviation from the average purchasing power of the whole city region. Blue cells of high significance are visible in the core city and the nearby suburbs. Peripheral concentrations of low-income households are located in western and north-eastern municipalities, as well as in the far south of the city region (in smaller numbers and with lower significance). Hot spots are located in western and eastern suburbs of the core city, in the northern town of Donauwörth and towards the east where the Augsburg city region overlaps with Munich’s comparatively affluent suburbs.

Interactive maps for all results can be accessed online.

https://ils-geomonitoring.de/maps/505/view#/ (16.08.2023).

The cold-spot maps were complemented by factsheets from the city regions and their municipal administrations, as well as by socio-economic statistics presented in Tables S1 to S6 in the supplements. Based on the interpretation of these results, the project team selected case study regions for stakeholder interviews from their interpretation of the results presented in the supplementary material.

The ‘micro-analysis’ approach to measuring poverty concentrations is based on an understanding of poverty beyond the material and income perspectives of absolute and relative poverty. Alongside income, the state of the local environment in terms of socio-economic opportunities and environmental pressure is also analysed. The sample concept for a data-driven implementation in the project context of this paper is illustrated in Figure 3. Data was purchased for selected municipalities where cold spot selection served as input for the stakeholder interviews, with variable selection based on an initial synthesis of multiple deprivation approaches in the literature (see, for example, the selection presented in Fairburn/Maier/Braubach 2016 and references to our own preliminary research in Section 2 of this paper). For the purpose of our research project, we identified three dimensions to systemize determinants of deprivation: (1) socio-economic household characteristics and housing costs, (2) public services and accessibility of the city-regional transport network, and (3) environmental pressures and resources. The indicators listed below the maps show potential candidates for measuring the dimensions before weighting is used to arrive at a combined index of disadvantaged neighbourhoods (the map on the right). The differing geometric representations of input data (e.g., point locations for amenities and housing costs vs. areal representations of accessibility isochrones or noise bands) require rules-based transformations into target geometries like grid cells. Such procedures are readily available in GIS software packages. Ideally, the spatial granularity of results is sufficiently detailed to represent local variations between individual plots of land and city blocks.

Implementation of such concepts, however, involves a range of challenges to spatial analysis, particularly on a local neighbourhood scale. Thus data protection laws in Germany prohibit the release of information allowing individuals to be identified and located. A further constraint is that the retrieval of aggregated information is highly dependent on the goodwill of cooperation partners and individuals backing the idea of establishing such research data infrastructures within city administrations. Even where this is the case, cooperation partners may decline consent to publish their figures. The supplementary material of this paper shows Düsseldorf as an example where a cluster analysis mapping the spatial variation of locational advantages and disadvantages was successfully completed and published. Endeavours to extend such cooperation projects to suburban communities in the Ruhr region have failed through not all cities seeing the need and/or having the capacity to support the research.

Alternatives such as small-area statistics available from private data vendors are mainly designed to support businesses with location-based information on customer potentials (Herter 2018). Their use for scientific purposes is theoretically possible but subject to cost constraints and acceptance of the fact that the methods used to produce the datasets are not fully disclosed. The RWI’s release of its purchasing power dataset, which we used for our cold spot analysis, represents a way to publish this data, but many prerequisites have to be met. Moreover, grid size is limited to 1x1 km. The use of such data remains problematic when wanting to maintain the highest scientific standards and have reliable base data for a continuous and consistent monitoring of neighbourhood deprivation.

Mapping out data-driven concepts of deprivation potentially stigmatizes locations exposed to such socio-economic pressures. Critical geographers point to social constructivism theory, which indicates that results should be handled with care so as not to interfere with the research subject. This extends to cartography where mapping requires sensitive handling and sometimes a conscious decision not to publish results where a potential negative impact on an area’s image may fuel deprivation dynamics (Kühne 2021). At the same time, performing such analyses for social policy action requires exchanges among scholars on data potentials and methods (Martínez/Pfeffer/Baud 2016).

Against this background, neighbourhood deprivation analyses (‘micro-analyses’) could only be performed in an explorative manner for selected cold spots in the case study regions where a budget for purchasing data was provided by the project-funding party (in our case, the German Research Foundation). The author team has agreed not to publish location-based results for case studies in selected cold spots to avoid the neighbourhoods in question being stigmatized as deprived. Instead, Figure S1 in the supplementary material shows the sample implementation for Düsseldorf mentioned above.

The methods discussed in this paper to spatially analyse poverty concentrations are based on three main strands of poverty concepts. The material poverty perspective focuses on households and people unable to cover basic material needs with their income or savings. In the German context, they therefore qualify for social benefits including either the provision of social housing or financial support for housing on the free market. Methods to operationalize this concept are limited to the spatial analysis of benefit recipient concentrations at city and municipal levels. This approach is frequently used due to the ready availability of data. Results are therefore frequently reproduced in political debates on social policy and the governance of social and territorial disparities. However, the resulting information is not sufficiently representative for research questions aimed at analysing the displacement and relocation pressures affecting low-income households. Recipients of benefits are by and large shielded from gentrification dynamics within city limits in German city regions, making it unlikely that they will be forced to relocate to suburban areas where housing is cheaper, at least not in larger numbers. Instead, stakeholder interviews in the case study regions and scholars invited to discuss these results point to segregation pressures where local pockets of poverty emerge with no clear locational pattern along the core-to-suburban continuum.

See also the detailed reports on interview results of the project in Pfaffenbach/Dobusch/Weck et al. (2023).

This observation prompts a need for additional analysis methods targeting low-income households with incomes just above the thresholds for receiving benefits. In public debates on poverty and spatial justice, awareness is growing that the dynamics of socio-economic disparities are putting the ‘working poor’ at risk of poverty. They are the ones especially vulnerable to gentrification in the face of rising costs of living¸ with housing costs frequently the dominant cost factor. Locational concentrations of this group can be pinpointed through mapping purchasing power cold spots at the city-regional level. Together with auxiliary data on housing costs, such mapping can be used to identify concentrations of households at risk of poverty and establish a baseline for future monitoring. An additional advantage of this approach is the higher data granularity of 1x1 km cells compared to the limitations of aggregated city and municipal data normally used for spatially analysing the receipt of benefits.

Finally, the neighbourhood deprivation concept zooms in further when analysing selected cold spots. For the areas in question, a multi-criteria framework of neighbourhood resources and pressures is used to compute an index of disadvantage in 100x100 metre cells. This approach is very demanding in terms of data availability. Data providers able to supply such data only exist in larger cities, meaning that suburban communities are rarely covered. Moreover, the data is subject to data protection laws prohibiting researcher access to data where individuals can potentially be identified. The approach can therefore only be used for small-area estimates, with auxiliary data sourced from private data vendors at high cost.

In the project context, data was obtained for three municipal case studies at a total cost of approximately 16,000 euros.

The combination of these methods represents a comprehensive chain for analysing concentrations of low-income households in German city regions. The main contribution of this paper is the exploration of data-driven implementation options for the poverty concepts that we cover in this paper. As a methodological building block, they contribute to a wider research project extending to research questions on the monitoring of displacement trends internationally known as ‘suburbanization of poverty’. Despite the findings presented in this paper, limitations remain. First, the question guiding the research was whether there is pressure to relocate from attractive inner-city locations to the suburban fringe in Germany, similar to that observed in the United States and the United Kingdom in the aftermath of the financial crisis. The data-driven methods presented in this paper only allow poverty concentrations to be analysed, but not movements across the administrative borders of cities and suburban municipalities.

In this respect, the interviews conducted with stakeholders and housing experts in the case study regions support the finding that households on benefits are shielded from relocation pressures within city limits. Even if gentrification is leading to rising housing costs, thus squeezing out low-income households, city administrations are still by and large successful in providing housing alternatives within their jurisdiction, at least for households on benefits. However, the material poverty method has little to say about the displacement pressures faced by the working poor.

This methodological deficit cannot currently be overcome by the presented cold-spot analysis (i.e., the relative poverty method) either. Despite its general suitability for detecting poverty concentrations and their spatial variations over time, it is not suitable for attributing new poverty concentrations to displacement or relocation pressures. When presented with the maps, interviewed stakeholders unanimously argued that, while the cold spots were remarkably accurate for detecting poverty concentrations, most of these concentrations had emerged in situ. Reasons mentioned point to the declining socio-economic situations of older residents in increasingly derelict housing estates and to image-related neighbourhood disadvantages. The role of displaced new arrivals was seen as marginal.

This finding is not surprising if one acknowledges that data-driven spatial observation cannot be expected to identify causal relationships between poverty concentrations and the forces driving the displacement and relocation of low-income households. It is exactly for this reason that the authors favoured a mixed-methods approach within an interdisciplinary research project, including expert workshops and stakeholder interviews, from project outset. The information gained from the respondents was not only used to validate the results of the cold-spot analysis, but also to guide the selection of indicators and the dimensions used in the micro-analysis conceptually presented in this paper.

We encourage readers to consult detailed project publications that draw on expert advice and stakeholder feedback on the factors explaining disadvantaged neighbourhoods and the related poverty concentrations (Weck/Dobusch/Pfaffenbach et al. 2023, Pfaffenbach/Dobusch/Weck et al. 2023).

This paper presents data-driven methods from Germany aimed at examining a common trend in Global North countries where housing in attractive inner city areas can increasingly only be afforded by affluent sections of society. The hypothesis of a ‘suburbanization of poverty’ originates from observations in the United States and the United Kingdom where persisting urbanization has led to new concentrations of lower-income households in suburban communities. Especially after the financial crisis in the late 2000s, the majority of these households apparently relocated from modernized inner city neighbourhoods to locations in the suburbs where housing remained affordable. The lack of neighbourhood resources and exposure to locational disadvantages in the suburbs potentially reinforces disparities between inner cities and the suburbs.

Based on a project funded by the German Research Foundation, spatial planners and human geographers combined forces to analyse whether this trend was to be seen in German city regions. The findings of this paper are a building block in this research. Three methods were found to be worthwhile examining, each representing a different conceptualization of poverty. One method that is very prominent in the German context operationalizes the concept of material poverty through a spatial analysis of households on benefits. A second one identifies purchasing power ‘cold spots’, with the results showing where ‘working poor’ neighbourhoods are relatively poorer than nearby neighbourhoods and the city-regional average. The last method conceptualizes neighbourhood deprivation as a lack of resources and opportunities as well as exposure to environmental and social pressures.

Selected results were discussed in expert groups and presented to local stakeholders in the city administrations of the case study regions. The respondents unanimously endorsed the ‘cold spot’ method as suitable for monitoring poverty concentrations. However, the experts could not agree on the ‘suburbanization of poverty’ trend being the cause of new poverty concentrations, instead observing increasing segregation forces within their jurisdictions. Housing markets seem to play a dominant role in reshaping territorial disparities in this respect. However, further research is needed to test whether such observations are biased towards the situation of households on benefits. This might be the case when stakeholders focus on the concerns of benefit recipients as the ‘clients’ of their daily work. It is also possible that respondents were not able or willing to speculate on future risks in their current assessment. In this context, current economic crises and the predicted shortage of affordable housing in German city regions are likely to exert increasing pressure on low-income households to relocate to affordable locations. One can therefore not rule out that the factors driving the suburbanization of poverty are yet to prevail.

For this reason, it seems prudent for the spatial sciences to establish monitoring concepts tailored to the observation of poverty concentrations and the related socio-economic disparities. The analysis chain presented in this paper, accompanied by dedicated surveys and interviews, is thus considered as contributing new methodological knowledge in this respect.