In trying to figure out the best way to deal with America's too-high levels of diversity, we have trouble communicating to one another because we tend to only know our own neck of the woods, and maybe a handful of other areas.
That allows us to talk about blacks (or Mexicans) in terms of their sheer numbers as well as their share of the population, since those factors are not tied down to the physical space of my town, your town, or his town.
The upshot is a no-brainer, bearing in mind Robert Putnam's research on the corrosive effects of ethnic diversity on trust: greater percentages of blacks and Mexicans threaten homogeneity, sowing the seeds of distrust, bringing the place closer to being a shithole. The practical implications are also obvious -- keep their numbers and percentages down -- but harder to do much about here and now, such as preventing immigration and transplanting.
There are a host of other factors, though, that are more malleable in the short-to-medium term, and less likely to sound the alarm bells of "that's racist!" at a national level, where the federal government might get involved. These are spatial factors of residence, which only folks in the area would pick up on.
They also show a striking level of variety around the country. We should study these spatial patterns and see how they're linked to patterns in the things we care about, like crime rates, trust, civic participation, racial tension and riots, and so on. After all of these natural experiments, we can move away from the failed models in the future and try to re-shape them where they have already been carried out.
What are these spatial factors? From the maps I've looked at, there appear to be three:
1) Density. How densely packed or sparsely spread-out are people within a fixed boundary (neighborhood, "part" of a city, entire city, metro area).
2) Mixture. How much do the two (or more) groups overlap in territorial boundaries? The two could be separated, the smaller group's territory could be mixed within the larger group's territory, or two groups of roughly equal size could occupy the same territory.
3) Enclosure. In cases where the two groups are fairly separated, what is the shape of their boundary? I idealize this as four degrees of enclosure: 0-degree, where the boundary is a straight line; 1-degree, where the line is bent or curved at one point, forming an open alligator mouth; 2-degree, where the line is bent at two points, forming a U shape; and 3-degree, where one group is surrounded by the other.
To see how these spatial patterns vary across the country, check out this series of maps by Eric Fischer, drawn from 2010 Census data. (See this related series, using 2000 data. The list of cities is mostly the same, although not entirely, so check the other list if you don't see a city you're interested in.) Each dot shows 25 people of a given race: whites are red, blacks are blue, Hispanics are orange, Asians are green, and Others are yellow. There are maps for over 100 cities large and small across the country.
The problem in studying them is that it is easy to tell how each city scores on the three spatial variables, but requires digging through other data sources to link them to a crime rate, level of racial tension, history of race riots, and the like. We're mostly going by our impressions from what we've heard about other places -- or whether we haven't heard much about them (no news is good news).
A research project could code each city for all the relevant variables, draw any associations, and give us the upshot. That'll take awhile, and others may be more interested in this than I am, so I'll just give some overall impressions of what spatial patterns are better or worse for civic society.
After this intro post, two follow-up posts will go into greater detail.