Mining Urban Performance: Scale-Independent Classification of Cities Based on Individual Economic Transactions

Stanislav Sobolevsky, Izabela Sitko, Sebastian Grauwin, Remi Tachet des Combes, Bartosz Hawelka, Juan Murillo Arias, Carlo RattiarXiv 2014 · OpenReview

Abstract

Intensive development of urban systems creates a number of challenges for urban planners and policy makers in order to maintain sustainable growth. Running efficient urban policies requires meaningful urban metrics, which could quantify important urban characteristics including various aspects of an actual human behavior. Since a city size is known to have a major, yet often nonlinear, impact on the human activity, it also becomes important to develop scale-free metrics that capture qualitative city properties, beyond the effects of scale. This paper proposes a novel approach of city scoring and classification based on quantitative scale-free metrics related to economic activity of city residents, as well as domestic and foreign visitors. Using anonymized countrywide records of bank card transactions from Spain, we demonstrate how scale-independent metrics reveal important properties of cities that complement patterns discovered with official socioeconomic statistics.

Introduction

Urban planners and policymakers traditionally rely on static, administrative statistics like census data or annual GDP to evaluate city health. However, these metrics are often outdated by the time they are published and fail to capture the dynamic, real-time behavior of people living in and visiting a city. Furthermore, comparing cities of vastly different sizes is difficult. Larger cities naturally have more economic activity, but this relationship is often nonlinear; due to "agglomeration effects," activity often grows faster than the population (superlinear scaling). To truly understand if a city is "performing well," one must look past its raw size and see if its activity exceeds what is mathematically expected for its population.

Methodology

The researchers proposed a three-step framework for scale-independent classification:

  1. Metric Extraction: Three key parameters were defined from bank data:
  2. Resident Activity (RA): Total transactions by city residents anywhere in the country.
  3. City Domestic Earnings (CDE): Total transactions made in the city by any Spanish resident.
  4. Foreign Attractiveness (FA): Total transactions made in the city by foreign visitors.
  5. Scaling Analysis: These parameters were modeled against city population ($N$) using a power-law relationship: $Y = Y_0 N^\beta$. If $\beta > 1$, the activity is superlinear, meaning it increases disproportionately with city size.
  6. Residual Analysis & Clustering: The "residuals"—the difference between a city’s actual activity and the theoretical value predicted by the scaling law—were calculated. These residuals represent scale-independent performance. Finally, k-means clustering was applied to these residuals to group cities with similar economic "signatures."
Figure 3: Scaling laws for Activity, Domestic Transactions, and Foreign Transactions across Spanish cities, showing both Large Urban Zones (LUZ) and Functional Urban Areas (FUA).

Data and Analysis

The study utilized a massive, anonymized dataset from Banco Bilbao Vizcaya Argentaria (BBVA) covering the year 2011 in Spain. This included approximately 4.5 million active Spanish residents executing 178 million transactions and 8.6 million foreign visitors executing 17 million transactions. Data was aggregated into two levels defined by the European Urban Audit: 24 Large Urban Zones (LUZs) and 40 Functional Urban Areas (FUAs). Official socioeconomic statistics (GDP and unemployment) were used for validation and comparison.

Results and Discussion

The "Bigger is Busier" Rule

All three economic metrics showed superlinear scaling. Notably, Foreign Attractiveness exhibited the highest exponent (ranging from 1.432 to 1.476 depending on the spatial unit), meaning that as a city grows, it becomes exponentially more attractive to international visitors compared to smaller towns.

Figure 6: Superlinear scaling of Foreign Transactions (exponent 1.476) for Large Urban Zones, illustrating the intense concentration of international economic activity in larger urban centers.

Scale-Independent City Profiles

The residual clustering revealed three distinct city profiles that traditional GDP data often misses:

  • Green (Over-performers): High scores across all metrics. This group includes major cultural hubs and tourist magnets like Barcelona, Toledo, and Santiago de Compostela. Interestingly, these cities often showed high unemployment rates, suggesting that economic vibrancy does not automatically solve deep-seated social issues.
  • Yellow (Mixed): Average or mixed performance. This group surprisingly included Madrid, suggesting its massive absolute wealth is "normal" once its massive size is accounted for.
  • Orange (Under-performers): Cities performing below their population-based potential, often located in remote areas or coastal regions with seasonal economies (e.g., Algeciras, Huelva). This suggests an over-reliance on seasonal tourism and a lack of year-round accessibility.

Conclusion

This methodology proves that transaction data provides a unique behavioral lens on urban performance. By removing the effect of city size, planners can identify the true economic signatures of cities, distinguishing between growth driven simply by population and growth driven by qualitative urban performance. Future work will involve applying this to other countries to determine if these economic signatures are universal or culturally specific.

Summary updated

Abstract

Intensive development of urban systems creates a number of challenges for urban planners and policy makers in order to maintain sustainable growth. Running efficient urban policies requires meaningful urban metrics, which could quantify important urban characteristics including various aspects of an actual human behavior. Since a city size is known to have a major, yet often nonlinear, impact on the human activity, it also becomes important to develop scale-free metrics that capture qualitative city properties, beyond the effects of scale. Recent availability of extensive datasets created by human activity involving digital technologies creates new opportunities in this area. In this paper we propose a novel approach of city scoring and classification based on quantitative scale-free metrics related to economic activity of city residents, as well as domestic and foreign visitors. It is demonstrated on the example of Spain, but the proposed methodology is of a general character. We employ a new source of large-scale ubiquitous data, which consists of anonymized countrywide records of bank card transactions collected by one of the largest Spanish banks. Different aspects of the classification reveal important properties of Spanish cities, which significantly complement the pattern that might be discovered with the official socioeconomic statistics.
BibTeX
@misc{sobolevsky2014mining,
  author = {Stanislav Sobolevsky and Izabela Sitko and Sebastian Grauwin and Remi Tachet des Combes and Bartosz Hawelka and Juan Murillo Arias and Carlo Ratti},
  title = {{Mining Urban Performance: Scale-Independent Classification of Cities Based on Individual Economic Transactions}},
  year = {2014},
  eprint = {1405.4301},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1405.4301},
}
Figure