PaperlayerView exact revision on arXiv

Source-linked AI summary

Inequality is rising where social network segregation interacts with urban topology

Gergő Tóth, Johannes Wachs, Riccardo Di Clemente, Ákos Jakobi, Bence Ságvári, János Kertész, Balázs Lengyel

arXiv:1909.11414v1physics.soc-phcs.SI

TL;DR

Prior big-data research had not tested whether social-network segregation relates to economic inequality. Using online-network and geographic data from Hungarian towns, the paper finds that fragmentation is associated with inequality, and that urban form is related to fragmentation.

  • Problem

    Prior big-data studies of social networks had not tested whether social segregation relates to economic inequality, despite networks’ potential role in access to resources and information.

    Source: Introduction · p. 2

  • Method

    The authors analyze an online network of about 2 million people in Hungarian towns, relate fragmentation to income inequality, and test a three-measure urban-topology index against social indicators.

    Source: Introduction · p. 2

  • Results

    Social-network fragmentation is positively associated with income inequality, and fragmentation plus existing inequality predict future inequality growth; distance, barriers, and concentrated amenities accompany greater fragmentation.

    Source: Discussion · p. 7

  • Takeaways & Limitations

    Urban planning is a potential long-run intervention point because the physical arrangement of residential areas is connected to social-network fragmentation.

    Source: Discussion · p. 7

    The analysis cannot establish causal effects of average distance from town center or physical barriers on network fragmentation, though the authors judge reverse causality unlikely.

    Source: Results · pp. 4–5

Abstract

from arXiv · show

Social networks amplify inequalities due to fundamental mechanisms of social tie formation such as homophily and triadic closure. These forces sharpen social segregation reflected in network fragmentation. Yet, little is known about what structural factors facilitate fragmentation. In this paper we use big data from a widely-used online social network to demonstrate that there is a significant relationship between social network fragmentation and income inequality in cities and towns. We find that the organization of the physical urban space has a stronger relationship with fragmentation than unequal access to education, political segregation, or the presence of ethnic and religious minorities. Fragmentation of social networks is significantly higher in towns in which residential neighborhoods are divided by physical barriers such as rivers and railroads and are relatively distant from the center of town. Towns in which amenities are spatially concentrated are also typically more socially segregated. These relationships suggest how urban planning may be a useful point of intervention to mitigate inequalities in the long run.

Introduction

The paper examines whether social-network fragmentation is related to income inequality and whether urban form helps explain differences in fragmentation across towns. It reports that fragmentation and initial inequality are associated with greater subsequent inequality, while geographic indicators predict fragmentation better than other social indicators.

Introduction

  • Prior big-data research had not tested the relationship between social segregation and economic inequality, despite proposed links between network structure, resource access, and inequality.Homophily and triadic closure are described as mechanisms that can produce macro-scale segregation.Source: Introduction · pp. 1–2
  • The study analyzes an online social network of about 2 million people across about 500 Hungarian towns, relating town-level fragmentation to income-inequality levels and changes from 2011 to 2016.Source: Introduction · p. 2
  • Urban space may help explain differences in segregation across cities: neighborhood sorting follows income or community, and policy choices shape cities more directly than social networks.Researchers have proposed improving access between neighborhoods as a practical measure to reduce extreme inequalities.Source: Introduction · pp. 1–2
  • A composite urban-topology index combines residential distance from town center, physical barriers dividing residential areas, and spatial concentration of amenities; each measure and the composite relate significantly to fragmentation.Source: Introduction · p. 2
  • Geographic indicators are better predictors of social-network fragmentation than other social indicators of segregation.Source: Introduction · p. 2
  • Income inequality rises more where social networks are fragmented and initial inequality is high; urban indicators relate to inequality through their relationship with fragmentation.The authors use the urban indices as instrumental variables in a model predicting economic inequality.Source: Introduction · p. 2

Results

Across 474 Hungarian towns, social-network fragmentation is associated with income inequality and with urban topology. The results also link instrumented fragmentation positively to inequality and show that its association with later inequality is stronger where initial inequality is high.

Results

  • Network fragmentation correlates with 2011 town Gini scores across all towns (r = 0.29) and in towns above 15,000 residents (r = 0.44).Source: Results · p. 3, Figure 1 · p. 3
  • Mean town Gini rose significantly from 0.474 in 2011 to 0.484 in 2016.Source: Results · p. 4
  • The interaction between 2011 inequality and network fragmentation is positively and significantly related to 2016 inequality, with fragmentation’s marginal effect increasing as initial inequality rises.The marginal effect becomes significant around the mean 2011 Gini.Source: Figure 1 · p. 3, Results · p. 4
  • Amenity concentration correlates positively with fragmentation (r = 0.253), but reverse causality cannot be ruled out because amenity locations depend on local purchasing power.Source: Results · p. 5
  • Distance from the center, physical barriers, and amenity concentration each have a positive, significant relationship with fragmentation, as does the composite index.Urban-structure measures outperform ethnic, religious, education, and political segregation measures in predicting fragmentation.Source: Results · pp. 4–6, Figure 3 · p. 6
  • In the second-stage 2SLS, instrumented social-network fragmentation is positively related to income inequality across the three urban-topology measures and their composite.The reported relationship is robust across those approaches.Source: Results · p. 6, Table 1 · p. 7
  • Model tests support the topology indicators as strong, reliable instruments, although the IV models fit better than OLS except for the SCA regression.Source: Results · p. 6

Discussion

The paper reports that social network fragmentation is associated with income inequality and that fragmentation and existing inequality predict future inequality growth. Urban distances, physical barriers, and concentrated amenities also accompany greater fragmentation, but the proposed causal chain and planning remedies remain unproven.

Discussion

  • Social network fragmentation is positively associated with income inequality, and fragmentation plus existing inequality predict significant future inequality growth.The paper says these relationships are based on large-scale evidence from cities and towns.Source: Discussion · p. 7
  • Towns with significant distances, physical barriers, or spatially concentrated amenities tend to have more fragmented social networks.Source: Discussion · p. 7
  • The authors cannot prove that poor urban design fragments networks and amplifies inequality; they propose urban planning as a possible intervention.They hypothesize that improving neighborhood access, mixing, and service distribution could help.Source: Discussion · p. 7

Materials and Methods

The study combines Hungarian social-network and town-level data with open geographic sources. It measures residential distance from town centers, barrier-driven division of residential areas, and the spatial concentration of amenities.

Materials

  • Materials include iWiW user locations and connections, town-level demographic and economic data, residential land-cover data, mapped barriers, and amenity locations.The iWiW data cover 2.8 million users and are governed by a non-disclosure agreement.Source: Materials · pp. 7–8

Methods

  • ADC estimates the average distance of residential points from a town’s center of gravity, with points allocated according to population and residential-area size.Small ADC values indicate compact settlements; large values indicate remote population fragments.Source: Methods · p. 8
  • SPB measures residential-area dispersion after rivers, major roads, and railroads divide residential polygons.Higher SPB values indicate barriers disconnecting larger fractions of residential areas.Source: Methods · p. 8
  • SCA quantifies the spatial concentration of amenities by clustering points of interest within a 500-meter radius and measuring cluster-size entropy.Higher values indicate amenities concentrated in a few clusters; lower values indicate more even distribution.Source: Methods · p. 8

Supporting Information 1: Gini coefficient as the measure of local income inequalities

The paper uses the Gini index to quantify income inequality from town income distributions, with income categories and town-year labels specified for the calculation.

  • The Gini index quantifies income inequality from town income distributions.For town i and year t (2011 or 2016), x_p and x_q are total incomes in categories p and q, and n is the number of income categories.Source: Supporting Information 1: Gini… · p. 12

Supporting Information 2: Online social network data and representativity

The study uses iWiW user profiles and friendship ties to analyze social networks across Hungarian towns. Its coverage was broad but varied by age and settlement size.

About the data

  • The February 2013 iWiW export provided basic user information and friendship ties for research.At its peak, iWiW had registrations from about 30% of Hungarians over 14; more than 4 million users listed a Hungarian settlement as their home location by early 2013.Source: About the data · p. 12, Representativity · p. 12

Representativity

  • By age, iWiW representation broadly followed the estimated online population until age 60, while its share of the total population peaked near age 30 and declined at older ages.The user-to-population ratio fell below 30% above age 50 and below 10% above age 70; the economically active population was well represented.Source: Representativity · p. 12, Figure 4 · p. 13

Potential biases in geographical representativity

Supporting Information 3: Principal components of urban topology indicators

The study combines urban-topology measures using principal component analysis (PCA), while also testing a version that excludes amenities because of possible reverse causality.

  • The Composite Urban Topology Index (CUTI) combines average distance from the center, physical barriers, and spatial concentration of amenities using PCA.Because reverse causality could not be ruled out for amenities, the authors also tested PC(ADC, SPB), which combines only distance and barriers.Source: Supporting Information 3: Principal… · p. 14
  • The first component explained 56% of variance across the three CUTI measures and 70% across the two measures in PC(ADC, SPB).The respective eigenvector values were 1.66 and 1.4.Source: Supporting Information 3: Principal… · p. 14
  • High values of CUTI and PC(ADC, SPB) indicate greater spatial segregation across the dimensions each index includes.These dimensions are distance from the center, physical barriers, and, for CUTI, concentration of amenities.Source: Supporting Information 3: Principal… · p. 14

Supporting Information 4: Controllers description, variable distribution and correlation

The supporting analysis describes social and demographic comparison variables and uses Random Forest variable importance to assess predictors of town-level network fragmentation. Urban-topology measures outperform the alternative drivers in that prediction exercise.

Supporting Information 6: Full tables of the 2SLS models

The supporting information lists 2SLS regression tables for fragmentation’s first stage and inequality’s second stage.

  • Table 3 presents fragmentation estimates from the 2SLS regression first stage.Source: Table 3 · p. 19
  • Table 4 presents inequality estimates from the 2SLS regression second stage.Source: Table 4 · p. 20
Loading 1909.11414v1…