1. Bibliographic Information

1.1. Title

Socio-spatial segregation and human mobility: A review of empirical evidence

1.2. Authors

• Yuan Liao
• Jorge Gil
• Sonia Yeh
• Rafael H.M. Pereira
• Laura Alessandretti

1.3. Authors’ Affiliations

• Department of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, Sweden
• Department of Architecture and Civil Engineering, Chalmers University of Technology, Gothenburg, Sweden
• Institute for Applied Economic Research (IPEA), Data Science Division, Brazil

1.4. Journal/Conference

Computers, Environment and Urban Systems (CEUS). CEUS is a leading, peer-reviewed journal in urban analytics, geographic information science, transportation, and complex urban systems, known for methodological rigor and cross-disciplinary relevance.

1.5. Publication Year

2025 (Published; the paper specifies Supplementary data DOI consistent with CEUS in 2025).

1.6. Abstract

The paper reviews empirical evidence on socio-spatial segregation—defined as the uneven spatial separation of social groups—and how it is shaped when we consider everyday mobility beyond residential areas (e.g., work, school, shopping, leisure). Traditional segregation research focuses on residential sorting; emerging mobility datasets since the 2010s (GPS traces, mobile phone data, social media) enable dynamic approaches capturing “activity space” and “experienced segregation.” The review addresses three questions: (a) strengths and limitations of integrating mobility data into segregation research, (b) how human mobility relates to residential versus experienced segregation, and (c) what key factors explain the link between mobility and experienced segregation. Synthesizing across themes of segregation, activity space, co-presence, and the built environment, the authors clarify core concepts, methodological differences, and provide actionable insights to reduce segregation.

1.7. Original Source Link

• Official PDF: /files/papers/693d2b6efab55b0207482a77/paper.pdf
• Publication status: Published in CEUS (Appendix C references “Supplementary data … doi:10.1016/j.cmpenurbsys.2025.102250”).

2. Executive Summary

2.1. Background & Motivation

• Core problem: Socio-spatial segregation has traditionally been measured through residential data (who lives where), but this misses the dynamic nature of daily life—people travel to multiple places (work, school, services, leisure), creating potential for diverse co-presence and mixing beyond neighborhoods.
• Why it matters: Segregation shapes access to resources, services, opportunities, and social cohesion. Understanding segregation in activity spaces can inform policies on transportation, housing, and urban design to reduce inequalities and promote integration.
• Gaps in prior research:
  1. Overreliance on static, residential measures.
  2. Insufficient understanding of how mobility patterns change segregation exposure.
  3. Methodological inconsistency across mobility-based studies (definitions, time scales, who counts as co-present).
• Entry point/idea: Provide a structured, cross-disciplinary review focused on human mobility and segregation, introducing a framework of three approaches (residential, built environment/network, activity space) and critically examining methods using emerging mobility data.

2.2. Main Contributions / Findings

• Conceptual framework: Three complementary approaches to measuring segregation—residential (static, area-based), built environment/network (potential co-presence via spatial accessibility and centrality), and activity space (dynamic, mobility-based).
• Methodological reflections: Identify inconsistencies in measuring activity spaces, defining co-presence (visitors vs residents), temporal resolutions (minutes vs daily aggregations), and limitations (correlational evidence, co-presence ≠ social interaction).
• Empirical synthesis:
  • Experienced segregation (across activity space) is often lower than residential segregation, but results vary by context, activity type (work vs leisure), urban form, transport accessibility, socioeconomic status, and ethnicity/birth background.
  • Mobility homophily (tendency to encounter similar groups) persists; lifestyle and preferences shape exposure to diversity.
  • Crises (e.g., COVID-19, disasters) amplify segregation by altering mobility and access.
• Explanatory factors: Activity demand & lifestyle, individual fears/trust and social networks, housing & urban sprawl, transport accessibility & equity, and urban design.
• Actionable insights: Promote mobility and accessible, inclusive urban design; integrate affordable housing with transit; address transport equity; design public spaces to foster diverse co-presence.

3. Prerequisite Knowledge & Related Work

3.1. Foundational Concepts

• Socio-spatial segregation: The uneven spatial distribution of social groups (by race/ethnicity, income, education, birth background) across geographic space, limiting inter-group co-presence and opportunities.
• Co-presence: Individuals being in the same place at the same time; a necessary precursor for potential interaction. Co-presence is not interaction itself but shapes exposure to diversity.
• Activity space: The set of places an individual visits in daily life (home, work, shops, parks), including the paths and durations involved. Contrasts with static residential perspective.
• Experienced segregation: An individual-level measure capturing how segregated one’s exposure is across all visited places over time.
• Visiting segregation: A place-based measure capturing how diverse the visitors to a specific location are over time.
• Built environment / network segregation: Potential co-presence structured by urban form, street networks, and transport accessibility (often measured via space syntax centrality and accessibility metrics).
• Mobility homophily: The tendency to move in ways that maintain exposure to similar groups, even outside residential areas.

3.2. Previous Works

• Classic segregation metrics: Dissimilarity Index, Exposure/Isolation, Concentration, Centralization, Clustering; spatially extended frameworks (Reardon & O’Sullivan, 2004; Massey & Denton, 1988).
• Space syntax: Network centrality (integration/closeness, choice/betweenness), visibility graph analysis—quantifying how urban form structures potential encounters (Hillier & Vaughan, 2007; Rokem & Vaughan, 2018).
• Activity space segregation: Early survey-based works (Wong & Shaw, 2011; Li & Wang, 2017; Park & Kwan, 2018) show that daily mobility can dilute or amplify segregation, depending on context.
• Emerging mobility datasets (2010s–): Mobile phone GPS, Call Detail Records (CDR), app-based location traces, social media (Twitter). Studies such as Moro et al. (2021), Nilforoshan et al. (2023), Östh et al. (2018), Xu et al. (2019/2024) quantify visiting/experienced segregation and encounter networks at high resolution.

3.3. Technological Evolution

• From static, census-based measures to dynamic, mobility-informed measures (big geolocation data).
• Integrations with network science and time geography (space-time prisms, encounter potentials).
• Emerging causal and counterfactual approaches are still rare; most studies remain descriptive.

3.4. Differentiation Analysis

• Compared to traditional residential analyses, this review:
  • Centers mobility as the lens to understand segregation.
  • Separates place-based visiting segregation (how mixed a location’s visitors are) from person-based experienced segregation (how mixed an individual’s exposures are).
  • Critically examines methodological differences across mobility-driven studies (spatial units, co-presence definitions, time resolution).
  • Bridges to built environment and transport equity literature to explain mechanisms and policy levers.

4. Methodology

4.1. Principles

• The paper is a systematic literature review focused on how mobility data reshape our understanding of socio-spatial segregation. It proposes a framework with three measurement approaches and synthesizes empirical results while reflecting on methodology.

4.2. Review Design and Data Collection

• Search design: Keywords spanning four themes—segregation, activity space, co-presence, built environment (see Table 1 in the paper)—combined in the query: titles/abstract/keywords include (1 AND 2) OR (1 AND 3) OR (1 AND 4).
• Source and selection: Scopus search (Oct 18, 2023), English-language journal/conference papers; 176 original articles plus reviews and select post-collection developments.
• Categorization:
  1. Segregation & Activity Space (mobility data-driven quantification).
  2. Segregation & Co-presence or Built Environment (network and design mechanisms).

4.3. Conceptual Framework: Three Approaches

• Residential approach (static): Co-presence is counted within residential areal units; segregation reflects unevenness across neighborhoods.

• Built environment/network approach (potential co-presence): Uses spatial network centrality and accessibility to infer where groups could mix given walk/car/transit connections.

• Activity space approach (dynamic): Uses empirical mobility data to quantify visiting segregation (place focus) and experienced segregation (person focus), often at high spatiotemporal resolution.

  The following figure (Figure 1 from the original paper) shows the three approaches schematically and how individuals’ mobility links residence to daily destinations:

  该图像是一个示意图，展示了社会空间隔离的三个关键方面：居民区（1）、建筑环境（2）和活动空间（3）。不同颜色的图标代表不同的社会群体和活动，强调移动数据如何影响个体的社会隔离水平。

4.4. Measuring Co-presence in Mobility Studies

• Who is co-present?
  • Visitors vs residents: Studies differ on whether a visitor’s co-presence is measured against residents of the area or other visitors present during the same time window.
  • The review prioritizes visitor–visitor co-presence for activity space segregation (closer to actual simultaneous presence in places).
• Temporal resolution:
  • Fine-grained (e.g., 5 minutes, 50 meters proximity) versus coarse-grained (daily aggregation).
  • Trade-off: Higher time resolution yields better co-presence fidelity but needs dense location data; many studies aggregate to day-level due to data sparsity.

4.5. Measuring Activity Space

• Spatial units vary:
  • Points of Interest (POIs) (e.g., restaurants, gyms).
  • Administrative units (e.g., census tracts).
  • Custom grids/cells (e.g., Voronoi polygons).
  • Network edges/segments.
• Implications:
  • Larger zones can mask micro-level segregation; high-resolution POI-level analyses reveal that adjacent venues may cater to distinct groups.

4.6. Segregation Metrics (Core Mathematical Forms Integrated)

Although the original review summarizes metric families rather than presenting explicit equations, many referenced studies rely on standard, authoritative formulas. Below are key metrics commonly used to quantify segregation; each formula is shown alongside its purpose and variables:

4.6.1. Dissimilarity Index (Evenness)

Concept: Measures how evenly two groups are distributed across areas.

Formula: $ D = \frac{1}{2} \sum_{i=1}^{n} \left| \frac{x_i}{X} - \frac{y_i}{Y} \right| $

Symbols:

• $n$: number of areas (e.g., tracts or grid cells).

• $x_i$: count of group X in area $i$.

• $y_i$: count of group Y in area $i$.

• $X$: total count of group X across all areas.

• $Y$: total count of group Y across all areas.

  Interpretation:

• $D \in [0,1]$; higher values indicate greater unevenness (more segregation).

4.6.2. Exposure/Isolation Indices (P*)

Concept: Probability that a member of one group shares an area with members of another (exposure) or the same group (isolation).

Formulas: $ P_{AB} = \sum_{i=1}^{n} \left( \frac{x_i}{X} \cdot \frac{y_i}{t_i} \right), \quad P_{AA} = \sum_{i=1}^{n} \left( \frac{x_i}{X} \cdot \frac{x_i}{t_i} \right) $

Symbols:

• $x_i$: count of group A in area $i$.

• $y_i$: count of group B in area $i$.

• $t_i$: total population in area $i$.

• $X$: total count of group A.

• $n$: number of areas.

  Interpretation:

• $P_{AB}$: exposure of A to B; $P_{AA}$: isolation of A with A.

• Values near 0 indicate low exposure; near group shares indicate higher exposure.

4.6.3. Moran’s I (Spatial Clustering)

Concept: Measures spatial autocorrelation (how similar values cluster across space).

Formula: $ I = \frac{n}{W} \cdot \frac{\sum_{i=1}^{n}\sum_{j=1}^{n} w_{ij} (x_i - \bar{x})(x_j - \bar{x})}{\sum_{i=1}^{n} (x_i - \bar{x})^2} $

Symbols:

• $n$: number of areas.

• $x_i$: value (e.g., proportion of a group) in area $i$.

• $\bar{x}$: mean of $x_i$.

• $w_{ij}$: spatial weight between areas $i$ and $j$ (e.g., adjacency or distance-based).

• $W = \sum_{i,j} w_{ij}$: sum of weights.

  Interpretation:

• $I > 0$ indicates clustering of similar values; $I < 0$ indicates dispersion.

4.6.4. Index of Concentration at the Extremes (ICE)

Concept: Captures concentration of populations at high vs low extremes (e.g., income).

Formula: $ ICE = \frac{N_{\text{adv}} - N_{\text{disadv}}}{N_{\text{total}}} $

Symbols:

• $N_{\text{adv}}$: count of population in the advantaged extreme (e.g., highest income quartile).

• $N_{\text{disadv}}$: count in disadvantaged extreme (e.g., lowest income quartile).

• $N_{\text{total}}$: total population in the area.

  Interpretation:

• $ICE \in [-1,1]$; -1 means entirely disadvantaged; +1 entirely advantaged; 0 indicates balance.

4.7. Methodological Limitations Identified

• Predominantly correlational studies; limited causal inference or counterfactuals.
• Co-presence does not equal social interaction; proximity may not translate to meaningful contact.
• Population biases in big geolocation data; uneven sampling (e.g., leisure-heavy social media, night-time app usage).
• Home/work detection and demographic inference are necessary but difficult to validate due to privacy/anonymization.

4.8. Mechanism Map: From Mobility to Experienced Segregation

• The review synthesizes five interacting aspects:

  1. Activity demand & lifestyle.

  2. Individual values, fears, trust, and social networks.

  3. Housing & urban sprawl (residence–destination structure).

  4. Transport accessibility & equity (ease of reaching diverse places).

  5. Urban design (form and function of public spaces).

     The following figure (Figure 2 from the original paper) visualizes how home/work/amenities and lifestyle-driven movements shape co-presence networks:

     该图像是示意图，展示了不同个体在活动空间中的关联及其与住宅、健身房、工作场所、餐厅等活动的连接。图中的颜色与线条代表了不同个体对活动需求和生活方式的影响，具体分析分布在各个章节中，如交通可达性（Section 6.4），住房与城市蔓延（Section 6.3）等。

5. Experimental Setup

5.1. Datasets (Across Reviewed Studies)

Because this is a review article, there is no single experimental dataset. Instead, the paper surveys studies using:

• Traditional data:
  • Travel surveys, interviews, census/register data (smaller samples; self-reported trips/activities).
• Emerging big geolocation data:
  • Smartphone GPS/app traces (meters/seconds resolution; months/years; millions of users).
  • Call Detail Records (CDR) (coarser; time-stamped tower connections).
  • Social media (geotagged tweets; leisure/non-routine bias).
• Auxiliary data:

  • Points of interest (POIs), land use, transport networks, and street network centrality (space syntax), to contextualize activity spaces and potential co-presence.

    Why suitable:

• High coverage and granularity enable dynamic measures (experienced/visiting segregation).
• Mobility records approximate co-presence; validations show correlations with social interactions.

5.2. Evaluation Metrics (Formal Definitions and Explanations)

Mobility-based segregation studies in the review rely on well-established metrics. Below are standardized forms and explanations (see also Section 4.6 for integrated methodology):

5.2.1. Dissimilarity Index

Conceptual definition: Quantifies unevenness in spatial distributions of groups across areas. Mathematical formula: $ D = \frac{1}{2} \sum_{i=1}^{n} \left| \frac{x_i}{X} - \frac{y_i}{Y} \right| $ Symbol explanation: See Section 4.6.1.

5.2.2. Exposure/Isolation (P*)

Conceptual definition: Probability that a typical member of a group encounters members of another/the same group in shared areas. Mathematical formulas: $ P_{AB} = \sum_{i=1}^{n} \left( \frac{x_i}{X} \cdot \frac{y_i}{t_i} \right), \quad P_{AA} = \sum_{i=1}^{n} \left( \frac{x_i}{X} \cdot \frac{x_i}{t_i} \right) $ Symbol explanation: See Section 4.6.2.

5.2.3. Moran’s I

Conceptual definition: Spatial autocorrelation measuring clustering/dispersion of a value (e.g., minority share). Mathematical formula: $ I = \frac{n}{W} \cdot \frac{\sum_{i=1}^{n}\sum_{j=1}^{n} w_{ij} (x_i - \bar{x})(x_j - \bar{x})}{\sum_{i=1}^{n} (x_i - \bar{x})^2} $ Symbol explanation: See Section 4.6.3.

5.2.4. ICE (Index of Concentration at the Extremes)

Conceptual definition: Balance of extremes (advantaged vs disadvantaged groups) within an area. Mathematical formula: $ ICE = \frac{N_{\text{adv}} - N_{\text{disadv}}}{N_{\text{total}}} $ Symbol explanation: See Section 4.6.4.

Note: Other metrics referenced (e.g., space syntax centrality, social interaction potential, i-STP, spatial segregation indices with distance-decay, Segregated Mobility Index) are conceptually described in the review but without explicit formulas in the paper. Where formulaic detail is critical in a particular cited study, readers are referred to the original sources for exact mathematical definitions.

5.3. Baselines

As a review, comparisons occur across methodological families rather than a single baseline vs. proposed model. Typical baselines include:

• Residential-only segregation measures versus mobility-informed activity space/experienced measures.

• Visitor–resident co-presence estimates versus visitor–visitor co-presence (more realistic for activity space).

• Coarse temporal aggregation (daily/weekly) versus fine-grained intervals (minutes).

  The following figure (Figure A.1 in Appendix A of the original paper) summarizes the distribution of reviewed studies by country and theme, indicating strong representation from the USA and China:

  该图像是条形图，展示了研究国家的数量和相关文献数量，包括a部分和b部分。a部分显示91篇文章中，涉及的作者数量为266，主要国家为美国和中国；b部分展示85篇文章中，涉及的作者数量为209，排名前列的国家为美国和中国。图中还展示了不同数据来源的文章数量变化趋势。

6. Results & Analysis

6.1. Core Results Analysis

• Is experienced segregation lower than residential segregation?
  • Frequently yes: Many studies find that daily mobility broadens exposure, reducing experienced segregation compared to residential segregation (e.g., workplaces, central amenities, public transport corridors).
  • But not always: Results vary; workplace segregation can be higher than residential in some contexts; leisure can be mixed depending on cost/culture; peripheral areas with poor accessibility show smaller differences.
  • Correlation persists: Residential segregation often correlates with experienced segregation (e.g., strong in some cities), yet correlations can be weak or even negative in specific subgroups (e.g., migrant exposure at home vs activities).
• Mechanism complexity:
  • Mobility homophily maintains exposure to similar groups.
  • Urban form and amenity distribution (diverse, central amenities) foster mixing; fragmented/peripheral forms hinder it.
  • Transport accessibility is a lever—inclusive networks reduce segregation; inequitable networks entrench it.
• Crises amplify segregation:
  • COVID-19 reduced public transport use, increased car reliance, deepening inequalities—especially for young and vulnerable groups in disadvantaged areas.
  • Disasters reveal mobility gaps (higher-income groups more likely to evacuate) and prolong post-disaster segregation.

6.2. Differences by Socioeconomic Status

• Wealthier groups (in developed contexts) tend to:
  • Travel farther and more often.
  • Visit diverse and dispersed activity spaces.
  • Experience lower segregation (more mixing).
• Less wealthy groups tend to:
  • Have constrained, localized activity spaces (limited affordability/access).
  • Experience higher segregation in activity spaces.
• Exceptions:
  • In some contexts, wealthy elites maintain segregated “flowing enclaves.”
  • In developing contexts, low-income groups may travel extensively due to urban form/job distributions.

6.3. Differences by Ethnicity/Birth Background

• Distinct mobility and activity patterns yield varied experienced segregation across ethnic groups.
• Leisure and non-routine activities show pronounced differences.
• Ethnicity interacts with socioeconomic status (intersectionality):
  • High-income natives and low-income minorities can both exhibit high experienced segregation.
  • For Whites, homophily often reflects preference; for Blacks/Hispanics, it is often constrained by accessibility.

6.4. Built Environment, Housing, Transport, and Design Factors

• Housing & urban sprawl:
  • Sprawl correlates with rising segregation; compactness improves job access and mitigates poverty segregation.
  • Spatial mismatch (residence far from jobs/services) increases commute burdens for low-wage workers and minorities.
• Transport accessibility & equity:
  • Inequitable networks reinforce segregation; equitable investments (fares, service coverage, multimodal access) can reduce experienced segregation.
  • Design biases (elevated highways, inequitable station siting) can displace or expose marginalized groups.
• Urban design:
  • Walkable, bikeable, mixed-use, high-integration environments foster co-presence and inclusion.
  • Parks and public spaces—if equitably accessible and inclusively designed—enhance social integration.

6.5. Policy-Relevant Insights

• Facilitate mobility to promote integration:
  • Integrate affordable housing with transit corridors.
  • Design inclusive public space networks (streets, parks, amenities) to support diverse co-presence.
  • Address fare affordability and service coverage for low-income/minority groups.
  • Plan for crisis resilience in transport/housing to avoid amplifying segregation.

6.6. Limitations of Evidence

• Co-presence captured by mobility data does not guarantee interaction.
• Strong dependence on data sources with biases (who carries phones, who posts on social media).
• Lack of causal inference limits policy translation; future work should incorporate counterfactuals and experimental/quasi-experimental designs.

7. Conclusion & Reflections

7.1. Conclusion Summary

• The review establishes a mobility-centric framework for understanding socio-spatial segregation, distinguishing residential, network/potential co-presence, and activity space approaches.
• Empirical evidence suggests experienced segregation is often, but not universally, lower than residential segregation—mediated by lifestyle, accessibility, urban form, and social preferences.
• Methodological choices (who counts as co-present, temporal resolution, spatial unit definitions) critically shape results; harmonization and transparency are needed.
• Actionable strategies emerge from built environment and transport equity research—integrating affordable housing near transit, inclusive urban design, and equitable transport policies can reduce segregation.

7.2. Limitations & Future Work

• Limitations acknowledged:
  • Predominantly correlational studies; co-presence ≠ interaction.
  • Data biases and validation challenges (home/work detection, demographics).
  • Fragmented methodological standards hamper cross-study comparability.
• Recommended directions:
  1. Integrate big mobility data with transport systems, land use, and urban form datasets to quantify how design choices affect experienced segregation.
  2. Bridge intended vs observed co-presence by incorporating empirical mobility into urban design evaluations.
  3. Pursue causal and counterfactual analyses (e.g., natural experiments, policy changes) to test how land use and transport interventions alter experienced segregation.

7.3. Personal Insights & Critique

• Transferability:
  • The framework is broadly applicable to other domains where exposure matters (e.g., health access disparities, educational opportunity landscapes, disaster resilience).
• Potential issues:
  • Over-reliance on geolocation as a proxy for social exposure may misinterpret contexts where digital interactions substitute or complement physical co-presence.
  • Without demographic ground truth, inferences (e.g., income quartiles via home location) risk misclassification; more robust privacy-preserving linkage methods are needed.
• Opportunities:
  • Combine mobility with surveys or ethnography to validate whether co-presence translates to meaningful interaction.
  • Use quasi-experimental designs around transport fare reforms, station openings, or park investments to establish causal effects on experienced segregation.
  • Develop standardized co-presence measurement protocols (visitor–visitor, time windows, spatial granularity) to improve comparability and policy relevance.