How, when and where do cities shrink? To start improving our understanding of shrinking urban areas, documenting the city landscapes that change is inevitable. Policymakers, planning officials, and urban designers have traditionally focused on urban urgencies occurring from expansion or growth. Consequently, models predicting the variety in which urban shrink unfolds are limited. In addition, as with any scientific modeling, the manipulation of variables to predict urban shrinkage is a simulation, a representation, of which the indicators have been empirically derived from past examples. The social shifts following the unification of post-World War II Germany or the ongoing industrialization of the agricultural complex and progress in GMO-technologies constitute only two examples from the second half of the last century of which the scale of effects on urban developments are uniquely complex to predict, model or map. The multidimensional patterns in human migration and settlement fluctuate drastically and present themselves particularly unpredictable in largescale disruptions that the global population is going to be confronted with during the twenty-first century alone––climate change, environmental warfare, dropping fertility rates, digitization of daily life, globalization versus glocalization, hyper individualism and the continuation of postmodern cultures and neoliberal policies. The models to assess urban shrink, therefore, are limited and in majority reactionary to the symptoms of shrink.

If we agree on the loose definition for urban shrink to be the continued loss of population in former population clusters, we can start aiming for more holistic mapping attempts; Inclusive of a variety of variables that may contribute to the change of demographics. Mostly, if we present a variety of indicators that touch upon causes and effects we can start expanding our knowledge on this theme and lift up the conversation. GIS operate on spatially oriented data. Therefore, the lack of data hinders informative maps. As a result, the opportunities to map urban shrink is limited by the available data and opportunities to collect new data. Luckily, given the theory on urban shrink and the availability of data, most of the indicators identified in academic literature as indicators and symptoms of shrink are available for the case study area, the Netherlands. These indicators include:

1. Age distribution of the population
2. Household type and size
3. Fertility rate
4. Regional in-/out migration
5. Labor market size
6. Size of labor market sectors
7. Vacancy of properties
8. Absolute and relative population growth
9. Social cohesion
10. Social participation

Therefore, this mapping project will present a set of municipality-level maps themed by different indicators for shrink. When combining the data and presenting different spatial levels of analysis, one can identify areas at potential risk for shrink. Through the spatial representation of population size projections, employment data, vacancy data, social data, demographic data, and topographic data, this research aims to model areas of risk for shrink in the Netherlands. In addition, this project presents a model to anticipate the shrinking areas of the Netherlands through an interactive online map. The online component enables to digest the large number of data variables at the level of each individual municipality. Also, the interactive map presents the outcome of the proposed model to predict urban shrink: all data variables synthesized into a score of higher or lower risk of urban shrink.

The urban structure of the Netherlands is different from the urban-rural structures of North America. Rather than several large metropolis areas in stark contrast to (disconnected) rural areas, the city-to-city and city-to-rural semilattice in the Netherlands is strongly developed through scale (the Netherlands is a country covering a small geographic area), historical relations (Dutch cities have historically specialized in public or economic functions instead of competed for similar functional roles), and because of the high-quality transportation networks (including roads for private transportation and infrastructure for public and alternative modes of transportation). As a result, the Netherlands does not have a single metropolis area with suburban agglomerations surrounding it and with a rural ‘rest’ of the country disconnected from the urban. Instead, there is a tight patchwork of population clusters spread throughout the country. The largest population clusters are titled in the map above. Yet, shrinking city trends might drastically change this urban-rural semilattice over time.

Natural population growth has historically been one of the main indicators for urban growth. A large household size used to be considered an ‘asset’, since children could contribute to intergenerational family wealth. The Millennial generation is the first to reverse this trend. A larger family size does no longer equate to more wealth. Instead, the increase of cost of living for urban residents makes children a financial burden instead of an asset. This map depicts fertility rates to be dropping in both urban as well as more rural areas.

Fertility rate data can be simplified in a two-option coding of either natural population decline or natural population growth. The mapping above depicts the scale at which the Netherlands is confronted with dropping fertility rates. Natural population growth is observed in strongly religious municipalities of the Netherlands only, areas belonging to the rural periphery or urban buffer zones.

The depiction of the aging population in the Netherlands for 2017 and for 2040 draws attention to the areas in dark red and black. These areas are likely to be confronted with shrink. Aging populations have special infrastructural and public service needs that require adaption of the urban space. The occurrence of people getting older is natural, undoubtedly; The relative increase of this age group within the total population is not, necessarily. The areas with a relatively large aging population include the border municipalities of the Netherlands, both in the east, the north, and the coastal south-west. A depiction of the aging population over time, overlaid with population clusters or urban cores enriches the narrative of demographic transition.

This animation depicts the demographic transition of the Netherlands between 2017-2040; Displaying the percentage of the population of 65+ relative to the total population. Overlaid with population clusters, one can observe areas outside main urban cores to experience high rates of aging, versus urban areas to remain largely ‘young’. More interestingly, some municipal areas with little aging are surrounded by a ring of rapidly aging municipalities. This phenomenon potentially implies the out-migration of elderly to tranquil, smaller towns because of a search for cheaper, safer, and easier living conditions for this age group outside of crowded, expensive, and fast-paced metropolis areas.

The relative change of the size of the labor market per municipality depicts worrisome conditions for a cluster of municipalities in the north of the Netherlands, mostly. In addition, border municipalities across the Netherlands depict significant shrink of the labor market. Implying a shrinking labor market, this visualization of spatial economic data speaks to the future relocation of residents. Historically, populations move to where employment opportunities are largest even despite automatization and digitization. However, other reverse trends of population migration have been witnessed in the past. During the mid-twentieth century, large-scale national improvement of infrastructure and the rise of car industry caused suburbanization processes in the US, for example. Yet, this process did not take place at a similar scale in the Netherlands.

For the regional balance of outflow versus inflow of residents, attention is initially drawn to a few red areas in the center-west of the Netherlands (where a higher number of residents left than arrived in 2018). These areas, when overlapped with the population clusters in the country, are the large urban areas of Amsterdam, The Hague, and Rotterdam. Indeed, this data might suggest a pattern of suburbanization. Yet, the areas in direct proximity to these large urban cores did not gain high numbers of residents arriving, which potentially suggests further, more remote spread. The departure of many residents in the large urban cores is related to the internationalization of these cities. Receiving increasing numbers of foreign visitors and settlers, these three cities have become less attractive for local residents to reside. In Amsterdam, for example, the dramatic growth of the tourism industry, combined with rising housing prices, pushes out residents. Again, other areas with a negative regional migration balance include some border municipalities.

Using the percentage of people joining local voluntary work per municipality in 2016, this map indicates the areas with the highest levels of social participation in dark green shades. Predominantly in the central east and north of the Netherlands, areas with higher community participation possess one of the more important assets in fighting urban shrink: a proactive community network. Unlike other maps, the municipalities further removed from large urban cores perform better on the variable of social participation. Larger urban areas tend to have high turnover rates of residents; Observing many residents move in and move out annually, estranging community members from one another. Such patterns affect social participation and weaken the community networks needed for a sustained quality of urban life.

In order to assess the level of social cohesion for the different COROP-regions, participants to the survey were asked “How high is your trust in…?” to each of the ten categories. Although a fluid concept, social cohesion is crucial component of any functioning society. The complexity of social cohesion as a concept is due to its relativity to different cultures, with intersections in place and time. Within economics, social cohesion can be reflected in social capita; Otherwise it can be related to sense of community, feelings of belonging and societal trust. The combined results from the different categories that were addressed in the COROP survey created a ‘trust score’. In 2018, the Netherlands overall had a 4.41/10 trust score. Compared internationally, this score of social cohesion is one of the highest in the world. Yet, the levels of institutional trust vary per region. Some border areas depict low levels of trust. Yet, it presents itself difficult to make further inferences about the visualized results.

In order to provide additional context to data about single-person households, this map has an overlay with population clusters. Some of these clusters are situated in areas with high rates of single-person households: Amsterdam, Rotterdam, The Hague, Groningen, Maastricht, etc. Most of these cities are attractive settling places for students and young professionals who tend to dedicate their time and resources to their career instead of to establish a family as early as previous generations would have. The scattered patchwork observed in the mapping does not take away the importance of understanding the role of single-person households within the shrinking city narrative. Areas under pressure of potential shrink, declining social participation or cohesion, and dropping fertility rates can face the effects of a downward spiral when the number of single-person households increases.

Complex economies of western countries like the Netherlands indeed have a large share of employment in highly educated labor sectors, like consultancy or IT. These economies are mostly situated in urban areas, as depicted in the mapping above. Areas with larger shares of manufacturing, industrial or agricultural jobs can be found in the east, north, and center of the Netherlands. These sectors are threatened by offshore production benefits, potentially causing urban shrink as witnessed in cities across the globe following deindustrialization (e.g. Ruhr-area in Germany or the Rust Belt in the US).

Above, the four maps complement one another in understanding the scope of residential property vacancy in Dutch municipalities. All data visualizations present themselves different. The last mapping is overlaid with population clusters. Then, one observes large areas suffering from high vacancy rates in the southwest, southeast, and some northern municipalities in the country. Moreover, combined with the insights from the mapping of long-term vacancies relative to the total number of vacancies (the third mapping in the series), concerns are raised for those areas with both generally high relative vacancy rates ánd above average high long-term vacancy rates.

Presenting itself as a final cornerstone variable within this research, the relative population growth per municipality in the Netherlands between 2017-2040 belongs undoubtedly in a model of future urban growth or shrink predictions. The results of this data visualization speak to an increasing rural-urban divide; Where urban cores grow and rural towns further shrink. Border areas in the southwest, southeast, east, and north of the Netherlands will suffer from significant population loss. Yet, this is data is derived from future predictions. Models are not reality. The variety of the preceding variables presented in this research are of equal importance within our understanding of changing urban patterns.

The different variables presented in this research are accumulated in a single large dataset, where the data are coded. Useful for its interactivity, the data display options available through symbols and popups, as well as accessibility for engagement; Online maps in Carto offer the opportunity to take the ten variables of this project to the next step; Establishing an urban shrink risk assessment score and index.

In the Carto mapping, one views the different municipalities in varying shades of red. The darker red, the higher the risk of regional shrink to take place in the area. This model is created through calculating the Netherlands’s general average for each category of data, accumulating the ten variables per municipality, and coding the municipal deviation from the national average. For example:

Regarding the demographic transition between 2017-2040, on average the portion of the population being 65+ of age will be 26.5% of the total population of the Netherlands in 2040. Municipalities that have a larger relative size of this age group in 2040 are coded with a ‘1’; Municipalities that have a smaller relative size of this age group are coded with a ‘0’. Through similar coding of the other data categories, a sum up score indicates the risk level of urban shrink in an area as measured across the ten different variables: where a score closer to 10 is an indication of potential shrink and a score closer to 0 is the reverse.

The calculation for the other variables is:

• On average, 38.2% of all households in the Netherlands are single person households in 2018. Municipalities that have a larger relative share of single-households within the total of all households are coded with a ‘1’; Municipalities that have a smaller relative size of this group are coded with a ‘0’.
• As soon as an area has an average fertility rate of below 2.1, the population will no longer grow from natural increase. Therefore, municipalities that have a fertility rate smaller than 2.1 are coded with a ‘1’; Municipalities that have a fertility rate equal to or larger than 2.1 are coded with a ‘0’.
• When the regional migration balance is smaller than zero, a municipality loses residents. Therefore, those municipalities are coded with an additional ‘1’. If the regional migration balance is larger than zero, a municipality can expect more arrivals than departures. Thus, a municipality is coded with an additional ‘0’ in that case.
• If the number of jobs between 2010-2018 increased within a municipality, it is coded with a ‘0’. On the contrary, if the number of jobs decreased it is coded with a ‘1’.
• The average size of the industry, agriculture, and manufacturing sector for the Netherlands as a whole is 15%, relative to the total size of the national labor market. If the industry, agriculture, and manufacturing sector uphold a larger share than 15% of the labor market in a municipality, it is coded with ‘1’. If the size of this sector within a municipality is smaller than 15%, it is coded with a ‘0’.
• 49% is the average number of residential properties in the Netherlands that have been long-term vacant. Municipalities that have more than 49% of their vacant properties belonging to the ‘long-term vacancy’-group are coded with a ‘1’; other areas are coded with a ‘0’.
• Between 2017-2040, the total population of the Netherlands is predicted to increase with an average of 5.98%. Municipalities of which the data indicate a growth less than 5.98% are coded with a ‘1’; municipalities of which the data show an increase of 5.98% or more are coded with a ‘0’.
• Regarding social cohesion, a score derived from the surveying about ten public institutions, both formal and informal, the Netherlands as a whole has an average score of 4.41. Municipalities that have a lower score of social cohesion than 4.41 are coded with a ‘1’, other with a ‘0’.
• Finally, 29% of all Dutch people join voluntary work, on average. Municipalities that depict a lower percentage of social participation through voluntary activities are given a ‘1’ score. Municipalities that have higher levels of social participation than 29% are coded with a ‘0’.

The summation of all these data entries can be at maximum a score of 10; Indicating a severe risk of shrink as measured by the variables. At a minimum, the score can be 0; Indicating the lowest possible risk of shrink as expressed through the selected variables.

If anything, this risk assessment index serves to open up dialogue about the indicators, causes, and related processes to regional and urban shrink. Through the establishment of any model, benchmarks need to be set and cutoffs are created. This puts limits on data and conclusions. Therefore, this model needs continuous improvement and contextualization. Yet, operating within an environment that lacks scholarly and policy attention to urban shrink, this model aims to situate shrinking cities at the center of urban design and planning and to demand holistic approaches when operating in these contexts.

In conclusion, Geographic Information Systems allow for the visualization of spatially oriented data. Within the field of study that is concerned with changing urban patterns – in this instance urban shrink – GIS enlarge our contextual understanding of demographic changes in cities by projecting the changing geographies relative to time, to place, and at a multiscale of topographies. Therefore, this section presents mappings that depict indicators and symptoms of urban shrink (yet, just a selection thereof). The aim of this visualization is to bridge between theory and spatial realities; Deepening our understanding of processes of urban shrink. When used as a model for future predictions, the maps inform where to situate policy interventions. What is more, the maps invite additional inquiry when exposing unexplained patterns or highlighting counter theoretical scenarios.

Link(s) to full project:

• Link 1 - thenewschool.carto.com/u/hijnr791/builder/ef450272-4512-4111-9329-ed3a4cbf360e/embed