MetaZipf. A dynamic meta-analysis of city size distributions

Area of Science:

• Urban studies
• Complexity science
• Geographic information science

Background:

• Urban scaling theory suggests city size distribution follows predictable patterns, offering potential for societal efficiency.
• However, significant variation exists in empirical findings regarding the slope of the rank-size distribution (Zipf's coefficient), hindering a unified understanding.
• Limited research has explored the diverse factors contributing to this observed unevenness in city sizes.

Purpose of the Study:

• To conduct a meta-analysis of Zipf's law to identify factors influencing variations in the Zipf's coefficient.
• To differentiate between technical, topical, and socio-demographic factors affecting urban size distribution.
• To investigate the temporal evolution of city size unevenness and its drivers.

Main Methods:

• A meta-analysis of 86 studies reporting empirical estimations of Zipf's coefficient was performed.
• An open database of 1962 coefficient estimates and associated metadata was compiled.
• Regression analyses, including dynamic meta-analysis, were used to model variations and evolution of the Zipf's coefficient.

Main Results:

• Technical choices in analysis specifications account for approximately 40% of the variation in Zipf's coefficients.
• Urbanization processes, rather than economic development or population growth, were found to be significantly associated with distinct evolutions in city size unevenness.
• No empirical evidence was found to support the effectiveness of past urban planning interventions in altering city size distribution patterns.

Conclusions:

• The heterogeneity in urban scaling literature is substantially driven by methodological and technical variations.
• Understanding urban size distribution requires focusing on urbanization dynamics over economic or demographic factors.
• Current urban planning strategies appear ineffective in modifying inherent city size distribution characteristics.