Understanding neighborhood income segregation around the clock using mobile phone ambient population data

Modern urban life is characterized by social and spatial separation based on socioeconomic status and other factors. While residential segregation is a well-studied phenomenon, understanding segregation beyond residential contexts, particularly its temporal dynamics, remains crucial. Existing research primarily focuses on residential segregation, neglecting the complexities of activity spaces and the temporal variations in social mixing throughout a day. This study aims to address this gap by examining income segregation in Guangzhou, China, using hourly ambient population data derived from mobile phone usage. The study explores how income segregation fluctuates throughout the day and across weekdays and weekends, and it investigates the influence of local urban functions and neighborhood contexts on these temporal variations. The findings contribute to a deeper understanding of human mobility, urban sociology, and the enduring nature of neighborhood effects, answering questions about social boundary crossing, the role of amenities and resources in driving urban mobility, and the long-term impact of short-term segregation dynamics.

Research on income segregation has traditionally centered on residential patterns, focusing on the development of segregation indices, the level and trends of income-based segregation across cities, and the consequences of such segregation on individual outcomes. Studies consistently demonstrate a rise in income segregation in recent decades, particularly in the US and increasingly in China. Consequences of residential income segregation often include disparities in access to resources, opportunities, and quality of life. However, a growing body of research recognizes that segregation extends beyond residential spaces, encompassing workplaces, leisure activities, and broader activity spaces. While studies have compared segregation in different contexts and examined individual-level experiences, research on income segregation in activity spaces and its temporal fluctuations remains limited. Previous studies have hinted at the temporal variation of segregation, revealing fluctuations throughout the day and differences between weekdays and weekends, but a comprehensive understanding of the underlying mechanisms remains elusive. This study builds upon this foundation by integrating temporal dynamics into the analysis of income segregation in activity spaces, focusing specifically on how urban functions and neighborhood contexts contribute to the observed patterns.

This study utilizes hourly ambient population data for eight income groups in 400m grid cells within the central urban area of Guangzhou, provided by China Unicom. The data spans a weekday and a weekend in June 2020. Income levels were classified by China Unicom using machine learning techniques based on various indicators. The local ordinal entropy index, a measure suitable for ordered groups, was used to quantify income segregation. Multilevel models were employed to assess the relationship between hourly income segregation and urban functions (transportation, institutions, residential, retail, accommodation, entertainment, offices) and neighborhood characteristics (number of residents, share of non-local migrants). Group-based trajectory analysis was used to identify distinct daily patterns of income segregation and to explore the stability of these patterns across weekdays and weekends. The analysis included sensitivity tests using simple information theory entropy and different numbers of income categories. Moran's I was used to assess spatial clustering of income segregation and group memberships. Multilevel logistic regression was performed to predict significant changes in group membership across days.

The study reveals several key findings. First, income segregation fluctuates significantly throughout the day and differs between weekdays and weekends, with higher variability observed on weekdays. The standard deviation of income segregation ranges from 0.366% to 18.425% across grids, indicating considerable heterogeneity in the extent of hourly fluctuations. Spatial clustering of income segregation is consistently higher at night than during the day, for both weekdays and weekends. Second, the presence of different urban functions and neighborhood characteristics significantly influences income segregation, exhibiting a strong temporal rhythm. Retail spaces (supermarkets, wholesale markets, street markets, shopping centers) are positively associated with higher income segregation, particularly during daytime hours. In contrast, accommodation and office spaces are associated with lower segregation, especially during daytime hours on weekdays. Neighborhood characteristics, such as the number of residents and the share of non-local migrants, also show significant relationships with income segregation, particularly during nighttime and morning hours. Third, group-based trajectory analysis revealed seven distinct daily patterns of income segregation for both weekdays and weekends, with most grids showing relatively consistent segregation levels throughout the day. There is a high degree of stability in group membership across weekdays and weekends, with less than 10% of grids experiencing substantial changes in segregation patterns. Spatial clustering is also observed in the distribution of these segregation trajectories. The presence of certain urban functions (schools, shopping centers, entertainment facilities) is negatively correlated with significant changes in group membership, while the share of non-local migrants is positively correlated.

The findings address the research question by demonstrating the dynamic nature of income segregation beyond residential contexts. The significant temporal variations in segregation, coupled with the influence of urban functions and neighborhood characteristics, challenge the notion that neighborhood effects are solely driven by static residential patterns. The study emphasizes the importance of considering both spatial and temporal dimensions when analyzing segregation. The consistency in segregation patterns across days, despite hourly fluctuations, highlights the enduring nature of neighborhood effects. The findings are relevant to urban planning and policy by illustrating how design choices and resource allocation can shape social interaction and potentially mitigate segregation. For instance, promoting mixed land use and diverse amenities could enhance social mixing and reduce segregation.

This study provides novel insights into the temporal dynamics of income segregation, highlighting the influence of urban functions and neighborhood characteristics. The findings demonstrate the enduring nature of neighborhood segregation and its real-time responsiveness to contextual influences. Future research could explore the causal relationships between urban functions and segregation patterns, examine segregation in virtual interactions, and investigate longer-term temporal trends and the impact of specific events. The limitations of the study should be considered when interpreting the results.

The study's main limitations include the correlational nature of the findings regarding the influence of urban functions, the potential for biases in the mobile phone data (underrepresentation of certain population groups, variations in phone usage), the focus on a single city, and the relatively short observation period. The lack of direct measures of social interaction and the reliance on a proprietary income classification algorithm also pose limitations.