More recent, and more sophisticated, “peer effects” research (by the likes of Carolyn Hoxby and Eric Hanushek) finds much the same. Rick Kahlenberg has been shouting from the rooftops that poor kids do better in “middle class” schools–which is why, in Gerald Grant’s words, there are no bad schools in Raleigh
I've heard of Spike Lee's "Magical Negro," but I've never heard of Petrelli's "Magical Rich Kid."
Now Hoxby's paper is under lock and key (Five dollars for an electronic download? Really?) and Kahlenberg and and Grant are merely spouting opinion. But Hanushek's paper is readily available. Here's the conclusion:
On average the black share of school enrollment in Texas is almost 30 percentage points higher for black students than for white students. Elimination of this gap would reduce the proportion black from roughly 0.39 to 0.16 for black students and raise the proportion black from 0.09 to 0.16 for whites. Using the coefﬁcient for blacks of 0.20 and the coefﬁcient for whites of 0.10, such a redistribution of students would reduce the racial achievement gap by 0.050 standard deviations in a single year. The cumulative effect of such a reduction for grades 5–7 (the sample period) depends upon the rate at which knowledge depreciates over time. If the rate of depreciation were equal to one minus the coefﬁcient on lagged achievement (roughly 0.4 for blacks and whites), the 3-year cumulative effect of racial composition equalization would reduce the race achievement gap by roughly 14%, moving it from 0.70 to 0.60 standard deviations.
These estimates represent extremes in the possible changes in racial compositions because they would require signiﬁcant changes in residences across districts and regions for blacks. More modest, and perhaps more achievable, changes still imply substantial closing in the test score gap.
So, at best we might possibly at the extreme see an 0.10 standard deviation increase in black student performance by sending your kid to an inner city public school.
Notwithstanding Hanushek's assertions to the contrary, a 0.10 standard deviation increase is student performance falls far short of an educationally significant result (0.25 standard deviation). Moreover, this kind of result, meager though it is, is only obtainable in the fantasy world of a data-mined economics study. In the real world, we're not going to see anywhere near such a large effect. Typically, educationally insignificant effect sizes don't show up at all in the real world. That's why we have the concept of educationally significant.
About the only thing you can be sure of by enrolling your pampered suburbanite children in an inner city school is that they will quickly learn how to catch a beating.
(Note: Even Checker laughs at this one.)