[Paper Review] The scaling of human interactions with city size

Markus Schläpfer, Luís M. A. Bettencourt+6 more|arXiv (Cornell University)|Oct 18, 2012

Complex Network Analysis Techniques46 references4 citations

TL;DR

This study analyzes nationwide mobile and landline call records from Portugal and the UK to show that both total communication activity and number of contacts scale superlinearly with city size, while local clustering of contacts remains constant. These findings reveal a scale-invariant acceleration in interaction-based spreading phenomena, providing empirical support for the role of human interaction networks in driving superlinear scaling of socioeconomic outcomes in urban systems.

ABSTRACT

The size of cities is known to play a fundamental role in social and economic life. Yet, its relation to the structure of the underlying network of human interactions has not been investigated empirically in detail. In this paper, we map society-wide communication networks to the urban areas of two European countries. We show that both the total number of contacts and the total communication activity grow superlinearly with city population size, according to well-defined scaling relations and resulting from a multiplicative increase that affects most citizens. Perhaps surprisingly, however, the probability that an individual's contacts are also connected with each other remains largely unaffected. These empirical results predict a systematic and scale-invariant acceleration of interaction-based spreading phenomena as cities get bigger, which is numerically confirmed by applying epidemiological models to the studied networks. Our findings should provide a microscopic basis towards understanding the superlinear increase of different socioeconomic quantities with city size, that applies to almost all urban systems and includes, for instance, the creation of new inventions or the prevalence of certain contagious diseases.

Motivation & Objective

• To empirically investigate how the structure of human interaction networks scales with city size.
• To test whether superlinear scaling of socioeconomic indicators stems from increased social connectivity per capita in larger urban areas.
• To examine whether local clustering of contacts remains constant across city sizes, despite overall network densification.
• To validate the role of communication networks in accelerating spreading phenomena such as disease or innovation diffusion.
• To provide a data-driven, microscopic basis for understanding urban scaling laws using large-scale mobile and landline communication data.

Proposed method

• Constructed reciprocal (REC) and non-reciprocal (nREC) interaction networks from mobile phone call detail records (CDRs) in Portugal and landline call records in the UK.
• Assigned individuals to urban units (cities, municipalities, urban zones) based on the geographic location of their most frequently used cell tower or exchange area.
• Analyzed the scaling behavior of total number of contacts and total communication activity (call volume and count) with city population size N.
• Measured local clustering coefficient to assess whether contacts of an individual are also connected to each other, and tested its constancy across city sizes.
• Used epidemiological models to simulate spreading dynamics on the inferred networks, confirming scale-invariant acceleration.
• Applied data filtering to exclude business hubs and non-personal calls, ensuring the networks reflect personal social ties.

Experimental results

Research questions

• RQ1 How does the total number of contacts per individual scale with city population size?
• RQ2 How does the total volume of communication (call count and duration) scale with city size?
• RQ3 Does the local clustering coefficient of an individual’s contacts remain constant across different city sizes?
• RQ4 To what extent do communication networks in larger cities exhibit accelerated spreading dynamics?
• RQ5 Can large-scale communication data empirically validate theoretical models of superlinear urban scaling?

Key findings

• The total number of contacts per individual and the total communication activity both scale superlinearly with city population size, with exponents β ≈ 1.15, indicating a multiplicative increase in social connectivity as cities grow.
• The local clustering coefficient—measuring the likelihood that two contacts of an individual are also connected—remains largely constant across city sizes, suggesting stable local network structure despite overall densification.
• The scaling of communication volume and contact count is robust across different urban definitions (e.g., municipalities, urban zones) and data sources (mobile and landline networks).
• Epidemiological simulations on the inferred networks confirm a systematic, scale-invariant acceleration of spreading processes in larger cities, consistent with empirical urban scaling laws.
• The results support the hypothesis that superlinear scaling of socioeconomic indicators (e.g., innovation, crime, disease prevalence) arises from the scaling of underlying human interaction networks.
• Despite potential biases in mobile phone data, the study confirms that call records provide a reliable proxy for personal social interaction networks at national scale.