The Schelling Segregation Model
Nobody in this simulation wants segregation. Every square just wants to not be too outnumbered on its own block. Watch what the neighborhood does anyway.
Controls
A square moves if fewer than this share of its occupied neighbors share its color.
The rules of the market
The clean experiment: tolerance 0%, turnover on, steering on. Nobody cares who lives next door. Watch the index anyway.
Red lines
Red districts are closed to Group B: no purchase, no rental, no exceptions. Residents already inside stay until they move, and then cannot come back. The lines bite hardest with everyday turnover on. Draw them or use the preset, run the market, then lift them and count the rounds until the map forgets. Decide whether it ever does.
"Record this run" saves the current settings and results to the table below, so you can compare experiments.
Press Run. Predict first: where will this end up?
What happens as the rounds go by
Show the numbers as a table
| Round | Same-color neighbors | Wanting to move | Segregation index |
|---|
Your experiments
| # | Tolerance | Group B share | Vacant | Rules | Rounds | Same-color neighbors | Segregation index |
|---|---|---|---|---|---|---|---|
| No runs recorded yet. Set the sliders, run it, then press "Record this run". | |||||||
Before you touch the sliders
Predict: set tolerance to 30%, meaning every resident is happy to be a minority on their own block, two-to-one against. Write down what you think the map looks like after 50 rounds. Then run it.
Things worth trying
- Drop tolerance to 0%. Nobody cares who lives next door. What is the segregation index?
- Walk tolerance up 5 points at a time from 0. Is the change gradual, or is there a point where the map suddenly flips?
- Set tolerance to 70%. Does the model reach a stable arrangement at all? What does it mean when it can't?
- Set vacant lots very low (2%) and very high (40%) at the same tolerance. Why does the housing supply change how fast segregation happens?
- Make Group B 20% of residents. Compare each group's experience: does the smaller group end up more or less clustered?
- Give the groups different tolerances. Does one group's pickiness segregate everyone?
- Set tolerance to 0%, then turn on everyday turnover and steering. Nobody has any preference about neighbors at all. What does the index do, and what does that tell you about where segregation can come from?
- Turn on unequal shopping at equal tolerances. Whose same-color average rises, and why, when neither group got pickier?
- Press Preset district (or draw your own red lines), run the market, and read the map against the lines.
- Segregate the map under red lines. Lift them. Count the rounds until the segregation index comes back down, and decide whether it ever does. That answer is the most important number on this page.
What the base model does not contain
The point of the model is what it leaves out. No resident here is a bigot: nobody refuses to live near the other group, nobody is willing to be the only one on the block. There are no landlords, no realtors, no banks, no zoning boards, no price differences, no income differences, no history. Segregation shows up anyway.
That result cuts two ways, and both belong in your write-up:
- It explains something real. Segregation does not require anyone to intend it. Aggregate patterns are not just individual attitudes added up, so "people aren't as prejudiced as they used to be" does not predict that neighborhoods will mix.
- It is not a description of American segregation. Real segregation was built with tools this model has none of: restrictive covenants, redlined mortgage maps, urban renewal, exclusionary zoning, steering by real estate agents, and enormous wealth gaps that determine who can buy where. A model showing that segregation can emerge without those things is not evidence that it did.
The switches under "The rules of the market" add stripped-down versions of exactly those missing tools, one at a time. Red lines are a one-rule stand-in for restrictive covenants, FHA underwriting maps, and the agents who enforced both: one group simply cannot buy in the marked districts. Steering is the documented practice of showing families listings near "their own kind." Unequal shopping stands in for wealth and information gaps. None of it is calibrated to any real city, and this is a mechanism demonstration, not a measurement. The point is to isolate what each rule contributes, alone and in combination, using the run recorder. One rule at a time, like a lab.
Write: pick one mechanism the model leaves out. In a paragraph, say how you would add it to the grid, what rule would change, and what you would expect to see. Then say what you would need to measure in the real world to tell whether your version or Schelling's is doing the work.
Schelling, T. C. (1971). "Dynamic Models of Segregation." Journal of Mathematical Sociology, 1(2), 143–186. · The segregation index is the demographers' index of dissimilarity: the share of one group that would have to move for every neighborhood to match the city's overall mix. Real U.S. metro areas run from about 0.3 to above 0.7. Two cautions worth teaching: scattering people at random does not give 0 (chance clustering puts the floor near 0.15, which is why the tile reports the random starting value), and the number depends on how big you draw the neighborhoods, a real problem in segregation research known as the modifiable areal unit problem. · SOC 1113 · Midwestern State University