Trying to make sense out of odd integration standards

January 25, 2003

When the Census Bureau sets out to measure how segregated American cities are, it uses a curious 50-year-old measure that comes up with the improbable conclusion that Boulder is more integrated then Denver, and Salt Lake City more than Detroit.

That's because the bureau's measure is something called the "dissimilarity index," a count of how many black people would have to move in order for them to show up in equal numbers in every census tract. Typically, that number is much lower in places that have hardly any black people, but having hardly any black people (1.2 percent in Boulder, for example) is not what most Americans have in mind when they describe a neighborhood as "integrated."

But it's the Census Bureau, and they are supposed to know about this stuff, so every year when the rankings come out they are duly reported - and, usually, lamented - because they're interpreted as an indication of lingering white resistance to integration. But you'd think someone who's consciously resisting integration is unlikely to choose Detroit as the best place to move to. Or Denver, which, because of the huge growth in the Hispanic population, is close to becoming a "majority-minority" city.

Lois Quinn and John Pawasarat, researchers at the University of Wisconsin at Milwaukee, have developed an alternative definition of black-white integration that, as they say, exposes "the biases and limitations of the segregation indexes." They looked at every one of the 8.2 million blocks in the 2000 census and said a block was integrated if at least 20 percent of its population is black and at least 20 percent of its population is white. The study is available at www.uwm.edu/Dept/ETI on the Web.

"The historic segregation index," they say, "appears to have a built-in bias suggesting that integration (or nonsegregation) is defined by what the majority will tolerate." Their measure "describes a more democratic ideal that suggests each racial group finds the other group acceptable as neighbors."

Their rankings are sharply different from the Census Bureau's, as you can guess, but they comport better with people's intuitive understanding. The Milwaukee area, for example, is the second or third most segregated in the country according to the usual index, but it is 43rd out of the largest 100 by Quinn and Pawasarat's measure. The city of Milwaukee proper is in the Top 10 of the 50 largest cities, with more than one of every five city residents living on an integrated block.

To meet the Census Bureau's new desegregation goal, which also aims at smoothing out population density, 89 percent of Milwaukee's black residents would have to move. You might reasonably ask, why are blacks the only people who are expected to move to forward desegregation? And what's the matter with high downtown densities anyway? That's supposed to be a good thing, an antidote to urban sprawl.

Because the Census Bureau's rankings deal only with black-white residential patterns, the researchers focused their analysis on that question to illustrate how different their method is from the usual one.

Even with that relatively narrow focus, it's not clear that 20 percent is the most appropriate number. Colorado's black population is only 3.8 percent of the state total, for example, and in Denver it is 11.1 percent. If there's any residential clumping at all as a result of black housing choices, and certainly you would expect some, then in most parts of the state there won't be enough blacks remaining to reach anything like 20 percent.

Also, to the extent this way of analyzing the problem proves to be useful at all, it would need to account for other racial and ethnic groups, especially Hispanics, since Colorado's Hispanic population is several times larger than its black population. Like other metro areas with a small black population, Denver ranks lower on the new index than the old, 71st rather than 50th, but omitting Hispanics underestimates how diverse Denver neighborhoods really are.

The best way to add other groups isn't readily apparent. Would you call a block integrated if no one group represented more than 80 percent of its population? That's the easiest way to bring multiple groups of different sizes into the discussion, but it has the disadvantage that it would identify a block that was, for example, half-black and half-Hispanic as integrated even though no whites at all lived there.

Perhaps there simply isn't a single formula that captures the reality of America's differently diverse cities. And that's a much nicer problem to have than the one the Census Bureau's outdated formula continues to find.