<- Back to laws

Agent-based model of emergent spatial sorting

Schelling Segregation Model

Even mild local preferences about neighbors can generate strong population-level segregation without any agent seeking the final pattern. It is a mechanism demonstration, not a complete theory of real segregation.

Scientific statusMathematical agent-based model
Predictive formLocal rule to emergent pattern
DomainSpatial sorting and coordination
EvidenceSimulation + empirical extensions
Key limitationAbstract agents and institutions
Common misuseSegregation proves mild preferences
INTERACTIVE MODEL

move if similar neighbors / occupied neighbors < threshold

Each agent inspects a local neighborhood and moves when its fraction of similar neighbors falls below a tolerance threshold. Repeated local moves can transform a mixed grid into clusters.

This live teaching simulation uses an 18 by 12 grid, two equal-sized groups, 12 percent vacancies, Moore neighborhoods, deterministic initial placement, and sequential relocation to a satisfactory vacancy. It demonstrates one mechanism; it is not a forecast of any real city.

54.8Simulated clustering index
(%)
0 %80 %
LIVE AGENT-BASED EXPERIMENTOne local rule. A citywide pattern.
Group A Group B Vacancy Dissatisfied
NEIGHBORHOOD GRIDROUND 00
Interactive visual model for Schelling Segregation Model.
0spatial position x18
COORDINATE TRACEINDEX / ROUND
Interactive visual model for Schelling Segregation Model.
clustering index dissatisfied agents

Blue and coral cells know only their eight adjacent positions. Outlined cells are below the selected similarity threshold. Run the model and watch local moves draw a macro-level curve.

CHANGE
Similarity threshold
WATCH
emergent clustering
MEANING
Local relocation decisions change the population-level clustering trace even though no agent selects the final citywide pattern.
VISUAL MODEL

Local discomfort can reorganize the whole map.

A mixed field separates as the similarity threshold rises. No square contains a plan for the citywide pattern; order appears through repeated moves and feedback.

mixed starting fieldlocal move ruleclustered outcome
01 / MEANING

What it actually says

Schelling built a family of spatial models in which agents care about the composition of a small neighborhood rather than the population as a whole. Agents who are locally dissatisfied relocate. A configuration can therefore become highly segregated even when many agents would accept an integrated neighborhood.

The important result is generative: a simple micro-level rule can produce a macro-level pattern that looks more extreme than any individual preference. Schelling also warned that the aggregate pattern does not reveal the exact motives that produced it. Similar-looking maps may arise from preferences, income, housing supply, exclusion, discrimination, networks, or several mechanisms together.

Compact formmove if similar neighbors / occupied neighbors < threshold
Best interpretationSpatial sorting and coordination evidence in emergence.
Important cautionAbstract agents and institutions.
"A useful law compresses a pattern. It does not erase the conditions that make the pattern true."
02 / ORIGIN

How the idea developed

The modern form emerged through observation, argument, and later refinement. The timeline separates the first insight from the version now used in textbooks and practice.[1]

19691969

Schelling circulates early models of separation and neighborhood tipping.

19711971

Dynamic Models of Segregation presents bounded-neighborhood and spatial proximity models.

19781978

Micromotives and Macrobehavior makes the micro-to-macro logic accessible across social systems.

TodayToday

Agent-based extensions add housing prices, income, geography, networks, discrimination, and empirically estimated behavior.

Historical cautionEponymous laws often change after their first publication. Popular wording may be broader and cleaner than the original evidence.
03 / MECHANISM

How the pattern works

The relation becomes useful only when its mechanism, measurement process, and operating range are visible.

01Local evaluation

An agent observes a bounded neighborhood, not the full distribution of the city or population.

02Threshold response

Below a tolerance threshold the agent seeks another location; above it the agent can remain.

03Feedback

One move changes several neighborhoods, possibly making new agents dissatisfied and triggering further moves.

04Path dependence

Vacancies, update order, initial placement, and search rules can lead identical preferences toward different final configurations.

MODELmove if similar neighbors / occupied neighbors < threshold

Each agent inspects a local neighborhood and moves when its fraction of similar neighbors falls below a tolerance threshold. Repeated local moves can transform a mixed grid into clusters.

04 / APPLICATIONS

Where it earns its keep

Applications are strongest when the law changes a decision, measurement, model, or experiment rather than merely providing an analogy.

COMPLEX SYSTEMS

Teach emergence from local rules

Application

The model makes it possible to watch micro decisions accumulate into a pattern no agent designed.

PROFESSIONAL NOTE

Treat the output as a mechanism demonstration and test alternative rules, not as a literal city forecast.

URBAN RESEARCH

Form competing spatial hypotheses

Application

Researchers can add prices, mobility constraints, discrimination, schools, amenities, and housing supply to compare mechanisms.

PROFESSIONAL NOTE

Calibration and validation against real moves are necessary before policy use.

PLATFORM DESIGN

Audit recommendation-driven sorting

Application

Repeated local choices in feeds, groups, or marketplaces can amplify homophily and reduce cross-group contact.

PROFESSIONAL NOTE

Digital similarity metrics are designed features, so platform rules belong inside the causal model.

05 / LIMITS & MISUSE

Where it stops working

The classic checkerboard removes most institutions that shape residential segregation: zoning, credit, wealth, transport, historical dispossession, steering, discrimination, housing construction, household structure, and unequal information. Its elegance is analytical, not documentary completeness.

A segregated outcome does not identify tolerant preferences as the cause. The model shows sufficiency under specified rules, not necessity in observed data. Results are also sensitive to neighborhood geometry, vacancies, agent ratios, relocation search, simultaneous versus sequential updates, and the definition of satisfaction.

Misuse

"Real segregation is accidental"

Better: Intentional exclusion, institutional constraints, and unequal resources can be central.
Misuse

"The model measures racism"

Better: Its abstract similarity preference is not a direct measurement of prejudice or discrimination.
Misuse

"Mild preferences always produce total segregation"

Better: Outcomes vary with thresholds, vacancies, geography, search, and initial conditions.
Misuse

"A final map reveals individual motives"

Better: Many micro mechanisms can generate similar macro patterns.
07 / REFERENCES

Sources and further reading

Original publications and serious secondary scholarship are prioritized over summaries.

  1. Schelling - Dynamic Models of SegregationThe primary 1971 paper presenting bounded-neighborhood and spatial models.https://doi.org/10.1080/0022250X.1971.9989794
  2. Schelling - Micromotives and MacrobehaviorBook-length development of unintended macro patterns from individual choices.https://wwnorton.com/books/Micromotives-and-Macrobehavior/
  3. Clark and Fossett - Understanding the Social Context of the Schelling Segregation ModelReview of the model, later evidence, and its relationship to residential segregation.https://doi.org/10.1073/pnas.0708155105
  4. Bruch and Mare - Neighborhood Choice and Neighborhood ChangeEmpirical and dynamic treatment of neighborhood preferences beyond a single fixed threshold.https://doi.org/10.1086/507856
CONTINUE EXPLORING

Related laws, with the relationship made explicit.

These are editorial connections, not claims that the laws are mathematically equivalent.

CONTINUE READING

Place this law inside the collection.

LAW 033 / 100 PUBLISHED