How can mild preferences segregate a whole city?

Because small preferences add up. In the 1960s and 70s the economist Thomas Schelling showed that people who only mildly prefer neighbours like themselves can still produce a highly segregated society, even if they would happily accept a mix up to some limit. He called it a general theory of tipping.

His experiment is simple enough to run at a kitchen table. Schelling used pennies and dimes on graph paper, one per square, standing for two groups. Any coin in an "unhappy" situation, with too few neighbours of its own kind, is moved. Each move changes the neighbours of the squares it leaves and joins, which can make other coins unhappy in turn. In his papers he showed that a preference that your neighbours be of the same colour, or even a preference for a mixture "up to some limit", can lead to total segregation. Malicious or not, the motives were indistinguishable in explaining complete separation.

The point is about emergence: what each person wants and what everyone ends up with can be very different. Schelling shared the 2005 Nobel Memorial Prize in Economic Sciences for work on conflict and cooperation through game theory.

The model has a limit, and it matters. It leaves out outside pressures such as Jim Crow laws. In real US cities, segregation was not just a matter of individual taste. Redlining, the practice of withholding financial services from neighbourhoods with many racial and ethnic minorities, and racial covenants shaped who could live where. So the model shows that taste alone could be enough, not that taste alone is what happened.