🔬 physics

Is segregation encoded in urban form? An entropy-based analysis

This study analyzes São Paulo to demonstrate that urban built-form entropy is non-linearly and asymmetrically linked to income segregation, revealing that high-entropy configurations are associated with distinct processes like elite enclaving and incremental development, while low-entropy forms reflect selective occupation shaped by planning and market forces.

Original authors: Vinicius M. Netto, Caio Cacholas, Camila Carvalho, Edgardo Brigatti

Published 2026-04-21

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Vinicius M. Netto, Caio Cacholas, Camila Carvalho, Edgardo Brigatti

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a city not just as a collection of people, but as a giant, complex puzzle made of buildings. This paper asks a fascinating question: Does the shape and arrangement of the buildings themselves (the "puzzle pieces") secretly encode how rich and poor people are separated from each other?

The authors, studying São Paulo (a massive city in Brazil), found that the answer is yes. The physical layout of the city isn't just a neutral background; it actively helps sort people into different neighborhoods.

Here is the breakdown of their findings using simple analogies:

1. The "Entropy" Meter: Order vs. Chaos

The researchers invented a way to measure how "messy" or "ordered" a neighborhood looks. They call this Built-Form Entropy.

Low Entropy (The Grid): Think of a classic chessboard or a military parade. Everything is in straight lines, aligned perfectly, and predictable. This is what you see in old city centers or planned suburbs.
High Entropy (The Jungle): Think of a dense forest or a chaotic bazaar. Buildings are scattered, streets wind around like snakes, and there is no obvious pattern. This is common in wealthy, exclusive enclaves with winding private roads, or in informal settlements where houses are built one by one over time.

2. The "U-Shaped" Surprise

Usually, we think that as a city gets "messier" (higher entropy), things might get more chaotic or segregated in one specific way. But the researchers found a U-shaped relationship.

Imagine a seesaw or a valley:

In the Middle (Low-Medium Entropy): The city is somewhat organized. Here, rich and poor people are actually less segregated. They mix a bit more.
At the Extremes (Very Low or Very High Entropy): Segregation spikes.
- The "Ordered" Extreme: Highly planned, grid-like areas often have strict rules (zoning) that keep lower-income people out. It's like a gated community with a very strict bouncer.
- The "Chaotic" Extreme: This is where the most interesting thing happens. The "messy" areas are actually the most segregated, but for two very different reasons:
  1. The Rich: Wealthy people build winding, complex neighborhoods with hidden streets and large, detached houses. It's like a maze designed to keep outsiders out. The complexity acts as a "soft wall" that makes it hard for strangers to navigate or see in.
  2. The Poor: In many parts of São Paulo, lower-income communities grow organically (informal settlements). They also end up with a "messy" layout because they are built incrementally without a master plan.

3. The "Soft Barrier" Analogy

The paper suggests that High Entropy acts like a "soft barrier."

If you are a wealthy person, you might choose a neighborhood with winding, complex streets because it feels exclusive. It's hard for a delivery driver or a stranger to find your house, and you can't easily see your neighbors. It creates privacy and separation.
If you are a lower-income person in an informal settlement, the "messy" layout is a result of necessity. Houses are built wherever there is space, creating a labyrinth that is tight-knit for the community inside but disconnected from the rest of the city.

4. The "Tipping Point"

The researchers found a specific "tipping point" in the messiness scale.

Below this point, the city is relatively mixed.
Once the city layout gets "messier" than this point, segregation explodes. The richer the area, the more "messy" (complex) the layout tends to be. The poorest areas also tend to be "messy," but for different reasons.

The Big Takeaway

The paper argues that urban form is not just a stage where segregation happens; it is an active actor.

Think of the city layout as a filter.

Some filters (like strict grids) filter people out based on rules and prices.
Other filters (like complex, winding mazes) filter people out based on confusion and lack of access.

In short: If you want to understand why rich and poor people live apart, don't just look at their bank accounts. Look at the streets. If the streets are a confusing maze, it's often because someone wanted it that way to keep people out, or because the system forced a community to build that way. The shape of the city tells the story of its social divisions.

1. Problem Statement

Residential segregation is a well-documented phenomenon, typically analyzed through demographic data (e.g., income distribution) and spatial indices (e.g., Gini, Moran's I). However, traditional approaches often treat the physical configuration of the built environment (urban form) as a neutral backdrop rather than an active component of segregation.

The Gap: Existing spatial measures capture where groups are located but fail to explain how the morphology of urban space (building footprints, street layouts, block structures) structures separation and interaction.
The Question: Does the configuration of the built fabric "encode" residential segregation? Specifically, are there non-linear relationships between Built-Form Entropy (BFE) and the spatial distribution/clustering of income groups?

2. Methodology

The study employs a multi-step quantitative framework applied to São Paulo, Brazil (Latin America's largest city), utilizing a regular tessellation to harmonize disparate data sources.

A. Spatial Framework

Tessellation: The city is divided into a regular 3×3 km grid (9 km² cells). This neutralizes administrative boundary biases (Modifiable Areal Unit Problem - MAUP) and allows for the overlay of building footprints and census data.
Data Sources:
- Built Form: Building footprints from OpenStreetMap and GeoSampa.
- Socioeconomics: Census tract data (2010 IBGE Census) regarding per-capita income.

B. Metrics Calculation

Built-Form Entropy (BFE):
- Concept: Adapted from Shannon's information theory. It measures the randomness/variability in the 2D arrangement of building footprints.
- Computation:
  - Each 3×3 km cell is converted into a binary matrix (built vs. unbuilt).
  - Block entropy ( $H_n$ ) is calculated for increasing block sizes ( $n$ ).
  - The entropy density ( $h$ ) is estimated by extrapolating $H_n/n$ to infinite block size using the ansatz $H_n/n \approx a + c/n^b$ .
  - Correction: A "corrected entropy" ( $h_c$ ) is derived to account for asymmetry in built/unbuilt ratios (e.g., large parks or rivers reducing entropy artificially).
- Interpretation: High BFE indicates irregular, complex, and unpredictable building arrangements; Low BFE indicates ordered, regular, and predictable patterns.
Residential Segregation Metrics:
- Income: Average per-capita income per grid cell (area-weighted from census tracts).
- Inequality (Gini Index): Measures income unevenness within each grid cell.
- Clustering (Moran's I): Measures spatial autocorrelation of income within each grid cell. High Moran's I indicates that similar income groups are clustered together locally.

C. Analysis

Statistical Modeling: The authors use piecewise (broken-stick) regression to identify non-linear relationships and thresholds between BFE, income, Gini, and Moran's I.
Spatial Analysis: Hot Spot Analysis (Getis-Ord Gi*) is used to identify statistically significant clusters of high and low entropy.

3. Key Results

A. Non-Linear Threshold Relationships

The study identifies a critical entropy threshold ( $h_c \approx 0.317$ ). The relationship between BFE and segregation is not monotonic but follows a distinct "U-shaped" or threshold-based pattern:

Below Threshold (Low Entropy): Income and segregation indices are generally moderate.
At Threshold: Income and segregation indices reach their minimums.
Above Threshold (High Entropy): Both income levels and segregation/clustering indices rise sharply.

B. Income Distribution Across Entropy Regimes

High-Income Groups: Heavily concentrated in high-entropy areas (above $h_c \approx 0.363$ ). These areas feature intricate morphologies (e.g., winding streets, detached housing, or high-rise fragmented fabrics) that act as "soft barriers," reducing intervisibility and through-movement.
Low-Income Groups: Also concentrated in high-entropy areas (just above the threshold), but driven by different processes: informal urbanization, incremental self-built settlements, and weak planning regulation.
Middle-Income Groups: More broadly distributed across the entropy spectrum, showing no sharp concentration.

C. Asymmetry in Segregation Mechanisms

The rise in segregation at the high-entropy end is steeper than at the low-entropy end.

High-Entropy End: Accommodates two distinct socio-spatial processes:
1. Elite Enclaving: Planned, high-design irregularity creating exclusivity.
2. Informal Settlements: Unplanned, incremental irregularity creating internal cohesion but external isolation.
Low-Entropy End: Reflects more ordered forms (e.g., historic CBD, regular grid systems) associated with planning and market filtering. Segregation here is present but less intense and driven by regulatory filtering rather than spatial intricacy.

D. Correlation of Indices

Gini and Moran's I: Strongly positively correlated ( $r \approx 0.59$ ). High income inequality within a cell tends to coincide with strong local clustering of similar income groups.
BFE and Segregation: High BFE is strongly associated with high Moran's I (strong local clustering), particularly in the high-income and mixed-income high-entropy zones.

4. Key Contributions

Integration of Morphology: Moves beyond treating urban form as a passive container. It demonstrates that Built-Form Entropy is a structural property that actively encodes and signals segregation.
Non-Linearity Discovery: Challenges linear assumptions in segregation studies by revealing a threshold effect. Segregation does not simply increase with disorder; it spikes at specific entropy levels, with distinct mechanisms operating at the high-entropy tail.
Methodological Innovation: Proposes a replicable workflow using Shannon entropy on building footprints, harmonized via a regular grid, to link physical form directly to social metrics (Gini/Moran's I).
Contextual Insight: Provides empirical evidence for São Paulo showing that both elite enclaves and informal settlements converge on high-entropy morphologies, albeit through different urban processes (planned distinction vs. unplanned accumulation).

5. Significance and Implications

Theoretical: Supports the view that urban morphology is an affordance and signal of segregation. High-entropy forms can function as "soft barriers" that limit exposure between social groups, reinforcing segregation without formal walls.
Policy & Planning:
- Suggests that planning interventions focusing solely on income mixing may fail if they ignore morphological entropy.
- Proposes that modifying block arrangements, permeability, and building configurations could alter the "soft barriers" of segregation.
- Highlights the need to distinguish between "ordered" segregation (planning/filtering) and "entropic" segregation (enclaving/informality) when designing interventions.
Future Research: The identified threshold ( $h_c \approx 0.317$ ) warrants testing in other cities to determine if it is a universal property of urban segregation or specific to São Paulo's socio-economic and morphological context.

In conclusion, the paper argues that segregation is encoded in urban form. The physical texture of the city—specifically its entropy—does not merely reflect social divisions but actively shapes the conditions under which they emerge, persist, and intensify.