How much does the picture of urban school systems I was painting yesterday change if you look at race as an indicator of socioeconomic disadvantage rather than family income? The answer is that even though studies show that race exhibits a substantial, income-independent influence on school achievement, the qualitative picture is similar if instead of looking at how cities do with educating poor kids we look at how they do with black kids:


As you can see here, nationwide performance by black children is pretty deplorable with 59 percent rating as "below basic." But given that unfortunate baseline, you can see that several urban school systems -- Charlotte, Austin, Houston, New York, and Boston -- are doing somewhat better than average, while others are doing quite a bit worse than average.

Now it's worth saying that this way of slicing up the demographics is no substitute for a proper regression analysis. You can really see this if you look, for example, at white test scores where suddenly the otherwise poor-performing DC public schools start looking good because the DC white population is unusually affluent and well-educated compared to whites nationwide. Unfortunately, the data available to talk about in order to get city versus city detail is a little sparse. The NAEP TUDA is the best way of getting apples-to-apples comparisons of different urban districts, but it doesn't cover very many cities and doesn't let you get nearly as rigorous as would be desirable.

Still, I believe the overall basic point holds. Demographic factors have a large impact on educational outcomes. When you try to control for demographic differences, suddenly big city school systems don't look uniformly dysfunction -- some are above average. But demographics aren't destiny -- some cities are doing much worse than others. Beyond that, on a nationwide basis poor children and non-white children are doing unacceptably badly even in "good" school districts.

We want to hear what you think about this article. Submit a letter to the editor or write to