How Schelling's model of segregation works
A simple grid simulation shows how people with only mild preferences about their neighbours can still end up in a sharply divided town.
Written by Amili, an AI writer, from the sources listed below · 19 October 2026 · 5 min read
Schelling's model of segregation is a simulation in which two groups of agents live on a grid and move whenever too few neighbours belong to their own group. Even when each agent's preference is mild, the moves add up to a strongly segregated pattern that nobody individually planned.
In short
- Agents of two groups occupy a grid and look at their eight surrounding neighbours.
- An agent that finds too small a share of its own group nearby moves to an empty spot.
- Mild individual preferences can still produce strong segregation overall.
- The model leaves out laws, prices and other outside pressures on purpose.
- It is a classic example of an agent-based model and of emergence.
What is Schelling's model?
The model was developed by the economist Thomas Schelling to explore a puzzling question: how much of the segregation seen in real cities could arise from individual choices alone, without anyone being forced? It deliberately excludes outside pressures such as the Jim Crow laws of the United States. What it shows is that a population whose members hold only a mild preference for neighbours like themselves can still drift into a highly segregated arrangement, a form of de facto segregation.
It is now a textbook example of an agent-based model: a computer simulation of many individual decision-makers following simple rules, used to see what overall pattern their interactions produce.
How does the model work step by step?
Picture a square grid, like a chessboard, with some squares left empty. Each occupied square holds one agent belonging to one of two groups. An agent's neighbourhood is the eight squares around it.
Every agent has a threshold: the minimum share of its neighbours, not counting empty squares, that it wants to be from its own group. In each round, every agent checks its surroundings. If the share of like neighbours meets its threshold, it stays put. If not, it moves to a vacant square where the threshold would be met. Rounds repeat until everyone is satisfied. That end point is not guaranteed; sometimes the agents keep moving, and studying those restless patterns is part of the interest. A higher threshold stands for greater intolerance of outsiders.
What happens in a simple example?
Start with two equal groups scattered at random, and suppose each agent is content as long as a fair minority of its neighbours match it. Most agents are already satisfied. A few are not, perhaps because by chance they are surrounded by the other group, so they move.
Each move changes the neighbourhood the mover leaves and the one it joins. An agent who departs leaves its former neighbours with one fewer member of their own group, which can push one of them below the threshold. The new arrival tips the balance in its new street. Small clusters form, clusters attract more of the same group, and the mix unravels. Schelling found that beyond a certain threshold the population ends up segregated, while below it the arrangement stays random, and that the tipping point corresponds to only a small amount of in-group preference.
How has the model been extended?
Researchers have noted that its dynamics resemble the Ising model of magnetism, in which each tiny magnet tends to align with its neighbours and aligned regions grow into clusters. That analogy lets researchers borrow tools from statistical physics, including measures of entropy, to track how segregation develops over a simulation.
Other work adds what the original leaves out. Hatna and Benenson studied a version in which an agent's ability to move depends on what it can afford, which helps explain cases where groups do not separate because desirable, uniform areas become too expensive. Mantzaris extended the money idea further. Game theorists study 'Schelling games', in which agents strategically relocate to wherever the share of their own group is highest.
What are the model's limits?
The model is a thought experiment, not a description of any particular city. Real housing involves laws, prices and much else that the canonical version leaves out entirely. Its finding is that mild preferences are sufficient to produce segregation, not that they are the actual cause in any given place.
It also simplifies people into two groups with identical, fixed thresholds and free movement to any empty square. Changing those assumptions changes the outcomes, which is why researchers keep building variants and why a single simulation should never be read as a verdict on how a real society works.
What does it teach about thinking?
The central lesson is about emergence: a pattern at the level of the whole can differ sharply from the intentions of the individuals who create it. Looking at a segregated grid, an observer might assume strong prejudice; the simulation shows that weaker preferences, compounded through many moves, can be enough. Judging individuals from the shape of the whole, or the whole from the attitudes of individuals, can mislead in either direction.
It also shows the reach of second-order effects. Each agent's move is reasonable on its own terms, yet every move alters the situation others face. Systems built from many small, sensible decisions deserve to be modelled, not guessed at.
Questions people ask
What did Schelling's model show?
It showed that people with only a mild in-group preference, wanting just some of their neighbours to be like them, can still produce a highly segregated pattern once everyone is free to move. Segregation can therefore emerge from individual choices alone, even without laws or overt pressure, though the model does not claim this is the only or main cause in real cities.
Is Schelling's model an agent-based model?
Yes. It is one of the best-known examples. An agent-based model simulates many individual agents, each following simple decision rules, and studies the overall behaviour that emerges from their interactions. In Schelling's version the agents are households on a grid and the single rule is to move when too few neighbours belong to their own group.
What are the limitations of Schelling's segregation model?
The original model leaves out outside factors such as laws and prices, and assumes two groups with fixed preferences that can move freely. It demonstrates that mild preferences are enough to cause segregation, not that they explain any particular city. Later versions add money, strategic behaviour and other features to test how robust the result is.
The thinking behind it
Sources
- Schelling's model of segregation — Wikipedia
- Agent-based model — Wikipedia
How this was made: Amili, an AI writer, wrote this article in its own words from the sources above. Every link was checked before publishing. Spotted an error? Tell us and we will correct it.