Inequality is rising where social network segregation interacts with urban topology

Growing wealth and income inequality pose significant challenges to development, economic growth, and social stability. While unequal access to resources and services contributes to this disparity, the interaction between structural factors like social networks and geography remains under-explored. This research examines how these factors interact to influence inequality. Social networks, while offering access to opportunities, can amplify inequalities due to homophily (similar individuals connecting) and triadic closure (friends of friends connecting), leading to social segregation. Geography plays a crucial role, as an individual's location strongly predicts their economic outcomes. Spatial constraints on social ties, especially across physical or administrative boundaries, limit access to resources and opportunities, influencing economic inequality through social network structure. This study investigates the interplay between these factors, analyzing an online social network in Hungary to assess how social network fragmentation and urban geographic features (distance from town center, amenity concentration, and physical barriers) contribute to income inequality.

Existing literature highlights the role of social networks in maintaining and amplifying inequalities. Homophily and triadic closure are key mechanisms driving social segregation by socioeconomic status. Studies show that geographic location is a significant predictor of future economic outcomes, with disparities observed even within small geographic areas. The local bias of social ties and reduced connectivity across physical boundaries reinforce the influence of geography on inequality. While prior research has linked social networks and geography independently to inequality, empirical studies examining their interaction are limited due to data challenges. This study addresses this gap.

The study utilizes data from iWiW, a Hungarian online social network, encompassing approximately 2 million users in around 500 towns. Income inequality is measured using the Gini index based on income tax data from the Hungarian Statistical Office. Social network fragmentation is assessed at the town level using the Louvain algorithm for community detection, calculating a fragmentation index (F) based on network modularity. Three urban topology indicators quantify spatial segregation: (1) Average Distance from the Center (ADC), measuring the average distance of residential neighborhoods from the town center; (2) Spatial Concentration of Amenities (SCA), quantifying the concentration of amenities in spatial clusters; and (3) Segregation by Physical Barriers (SPB), assessing the extent to which physical barriers divide residential areas. A two-stage least squares (2SLS) regression model is employed, with urban topology indicators as instrumental variables for social network fragmentation, to examine the relationship between urban topology, social network fragmentation, and income inequality. Control variables account for factors like population density, foreign direct investment, unemployment, and distance to the border. OLS regressions, falsification tests, and robustness checks are conducted to validate the findings.

The study reveals a positive correlation between social network fragmentation (F) and income inequality (G). Towns with higher fragmentation exhibit higher Gini coefficients. The relationship is dynamic, with the interaction of initial inequality and fragmentation significantly predicting future inequality growth. Higher initial inequality amplifies the effect of fragmentation on subsequent inequality. The three urban topology indicators (ADC, SCA, SPB) are significantly associated with social network fragmentation. Towns with higher ADC (greater average distance from the center), higher SCA (more concentrated amenities), and higher SPB (more significant physical barriers) exhibit greater social network fragmentation. A 2SLS regression model confirms that social network fragmentation, instrumented by urban topology indicators, is positively related to income inequality. Physical barriers (SPB) prove to be the most robust instrumental variable. Densely populated towns show lower inequality, while those near borders exhibit higher inequality. The urban topology indicators outperform alternative segregation proxies (ethnic, religious, educational, political heterogeneity) in predicting fragmentation.

The findings demonstrate a clear link between social network fragmentation, urban topology, and income inequality. The results support the hypothesis that geographic features contribute to income inequality by influencing social network structure. The dynamic relationship between fragmentation and existing inequalities highlights a self-reinforcing cycle. Policy implications are significant, as urban planning can influence social network structure and thus mitigate income inequality. The study’s robustness checks strengthen the findings, while acknowledging limitations in proving causal relationships.

This research highlights the critical role of social network fragmentation and urban topology in shaping income inequality. The dynamic interplay between existing inequality and social network fragmentation emphasizes the need for proactive urban planning. Future research could explore interventions to improve connectivity across neighborhoods, promote mixed-income housing, and foster a more equitable distribution of services, aiming to mitigate the negative impacts of social network segregation and improve economic outcomes.

While robustness checks were performed, the study cannot definitively establish causality between urban topology, social network fragmentation, and income inequality. Confounding variables might exist, and the long-term evolution of neighborhoods involves complex feedback loops not fully captured in this study. The data is limited to Hungary and the specific online social network iWiW, limiting generalizability. Reverse causality cannot be fully ruled out for all variables.