Table 3 Summary of our results in comparison with selected previous studies

Yamamoto [5]

{1973, 1975, 1990}

Plotting segregation indexes (location quotient)

Ethnic segregation of Turkish inhabitants in Berlin West (7 color-mapped areas)

In 1973, more than half of all Turks in West Berlin lived in Kreuzberg and Wedding. The research reports that in 1975 Turks were the most segregated compared to Germans, Italians, Greeks, Yugoslavs and other groups. By 1990, segregation between Turks and Germans had largely decreased.

Nakagawa [49]

{1965, 1970, 1975, 1980, 1985}

Hierarchical cluster analysis

Age-group segregation (the area was divided in a two concentric zone model of West Berlin)

The age groups 0-19 and 35 and over are more densely distributed in more densely in Outer Berlin than in Inner Berlin, and the age groups 20 to 34 tend to be more densely distributed in Inner Berlin.

Kemper [7]

{1991, 1995}

Plotting segregation indexes

• Ethnic segregation (2 zones, West and East, classified in a total of 6 colored areas)

• Age-group segregation (Comparison of West and East Berlin, classified in a total of 7 colored areas)

The study notes that age segregation was more pronounced in East Berlin before unification, while socio-economic segregation was more pronounced in West Berlin. After unification, there was a decrease in the age group of children under 6 in the former East Berlin. Also, segregation rates of the foreign population decreased in both former West and East Berlin.

Kröhnert and Vollmer [15]

{1992, 1994, 1996, 1998, 2000, 2002, 2004}

Cluster analysis

• Gender segregation (5 clusters, at country level)

Berlin is part of a cluster of German geographical areas segregated by gender in which “the sex ratio is above average (…) and the share of students is the second highest among all clusters. The cities have strong service and tourism sectors. Unemployment among young people is low. The proportion of people employed in service sectors is among the highest of all clusters” [15, p. 9].

Blokland and Vief [27]

{2007, 2012, 2016}

Plotting segregation indexes (location quotient)

• Ethnic segregation (5 color-mapped areas)

• Socio-economic (5 color-mapped areas)

Ethnic indicators:

• Foreigners (strong decrease)

• Persons with migration background (fair decrease)

• Migration background: Turkey and Arabic states (strong decrease)

• Migration background: European Union (stable)

Socio-economic indicators:

• Unemployed persons (stable)

• Long-term unemployed persons (stable)

• Non-unemployed persons receiving state subsidies (slight increase)

• Child poverty (slight increase)

Marcińczak and Bernt [28]

{2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019}

Regression trees; Hierarchical cluster analysis

• Ethnic segregation (7 clusters)

This research found the following clusters:

• Rising pluralist enclaves

• Non-isolated host communities I

• Stable pluralist areas

• Non-isolated host communities II

• Established and increasingly pluralist areas

• Stable non-isolated host communities

• Persistent host communities

Masías et al. [29]

{2020}

Multivariate Kernel Density Estimation; Non-Negative Matrix Factorization. Maps are provided for each dimension.

• Ethnic segregation (4 clusters)

• Age-group segregation (3 clusters)

• Socio-economic segregation (3 clusters)

Using a data science approach, it was possible to reveal highly interpretable patterns in the data, confirming the existence of the phenomena of ethnic segregation, age-group segregation and socio-economic segregation.

Present work

{2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019, 2020}

Multivariate Kernel Density Estimation; Dynamic Fuzzy C-Means; Maps and Bump charts are provided for each dimension

• Ethnic segregation (Changes from 7 to 8 clusters)

• Age-group segregation (4 clusters)

• Socio-economic segregation (3 clusters)

• Gender Segregation (3 clusters)

Macro dynamics

• The identification of a new cluster was determined.

Microdynamics

Migration background:

• Cluster 0: Lebanon and Turkey become since 2010 the most overrepresented in this cluster

• Cluster 1: The most overpreserented is USA

• Cluster 2: subpopulation from Iran is the most overrepresented

• Cluster 3: subpopulation from Poland is the most overrepresented

• Cluster 4: Only Germans are overreprested and the rest groups become underrepresented

• Cluster 5: Kazakhstan, Vietnam, former Soviet Republic, among others, are over-represented.

• Cluster 6: Spain, France and Italy, among others, remain overrepresented over time.

• Cluster 7: subpopulations with migratory backgrounds from Syria and China become the most overrepresent sup-population in the emergent cluster.

Age-group segregation:

• Cluster 0 only over-represents subpopulations the 65 and 90 subpopulations.

• Cluster 1 over-represents young adults, adolescents and teenagers

• Cluster 2 over-represents all groups, especially those aged 65-90

• Cluster 3 over-represents young adults and children.

Gender segregation:

• There is no residential segregation by gender. The clusters appear to mirror changes in population density across the city.

Socio-economic segregation:

• The clusters can identify areas where there is a higher density of people applying for unemployment benefits. Qualitatively, it can also be observed that there was a change in the distribution of these residential densities across the city in 2009 and 2020.