We are collecting town level data for the state of Connecticut and wanted to construct a measure of racial segregation in each town. I was planning to create a measure of non-white share of the population [(total population - white population)/total pop], and have run into a snag. As one might expect, not every town has enough racial diversity to trust the margins of error for the non-white population count. How have others handled this issue--when some towns have a reasonable margin of error while others are too large to trust?
Connecticut Voices for Children