The Richards paper, on the other hand, claims in the abstract that gerrymandering "generally exacerbates segregation". I wasn't sure how to reconcile this claim with Monarrez's paper, the graph from my comment, as well as https://mobile.twitter.com/alv9n/status/950388285318328321 (from the articles author).

So I read the Richards paper, and now I can't reconcile why it explicitly contradicts its abstract (and your claim) in multiple places:

> Interestingly, gerrymandering was inversely related to segregation, such that districts with higher levels of segregation were less gerrymandered than less segregated districts. ..

> Indeed, in the only direct examination of the effects of educational gerrymandering, Clark (1987) con- cluded that gerrymandering of attendance zone boundaries in Topeka in the 1950s and 1960s were either neutral or desegregative in nature.

The results of her paper show a minuscule mean effect on school segregation due to gerrymandering: between 0.001% (ie 0.00001)and 0.003% mean increase in the level of segregation, with an outlier of 0.006% for Asian-White segregation.

As I mention above, this contradicts the findings from the other papers discussed. But even if we take this paper as our source of truth, it still seems like residential segregation is responsible for the vast majority of district segregation, and that the total magnitude of the gerrymandering effect is tiny.