How arbitrary city boundaries skew our view of residential segregation

A new study in Nature Cities has examined how arbitrary boundaries affect the measurement of residential segregation in US cities. The research used advanced computational methods to test the reliability of traditional segregation estimates. Its findings carry important lessons for those studying urban inequality and social division. The study tackled a long-standing issue known as the modifiable areal unit problem, where artificial boundaries like Census tracts can distort population data. Researchers applied redistricting algorithms to create millions of alternative boundary maps for American cities. These simulations generated probabilistic distributions, offering a new way to assess segregation with uncertainty in mind.

The team combined urban demographics with realistic spatial criteria to ensure the alternative boundaries were practical. They discovered that in smaller cities, segregation estimates could shift dramatically depending on the boundary configuration. However, in larger cities, the figures stabilised, reducing the impact of boundary changes.

Contrary to some concerns, the study found no evidence that current Census tract-based estimates are systematically biased. This challenges the idea that existing methods wildly overstate or understate segregation levels. The findings suggest that while fixed boundaries can introduce variability, especially in smaller urban areas, they do not inherently skew results. The approach also provides a template for addressing similar errors in other fields like public health and economic geography. Researchers and policymakers are now encouraged to adopt complementary methods for more precise measurements.