Two groups live in a city. People form partnerships with those nearby. How much of the observed sorting is driven by where people live versus who they prefer?
Observed ethnic endogamy doesn't tell you what people prefer — it tells you who they paired with given who was around. In Swedish register data, residential proximity explains 20–40% of same-group partnering for non-Western groups, and ignoring it overstates endogamy for the most segregated groups while understating it for the most integrated. Studying social boundaries requires modelling the partner market itself: who is actually available, in a municipality, a workplace, a school. Without that, we risk reading geography as preference and mistaking opportunity structure for taste.