The key to this discussion, as Seth Masket wrote yesterday, is that redistricting involves trade-offs, with the biggest one involving squeezing out the most seats for the party compared with making each of those individual seats safer. Partisan gerrymanders — and yes, there were several of these after the last census — choose the goal of “maximizing party seats” over the goal of “keeping incumbents safe.”
Dave Weigel shows that in detail in his look at North Carolina’s partisan Republican gerrymander. North Carolina Republicans managed to flip three seats in their favor with a 2011 redistricting. They did that, however, by packing as many Democrats as possible into three seats — but also by making most of the GOP seats considerably less safe. NC-3, for example, went from having a 24 percentage point John McCain majority to only a 13 point McCain edge. Sure, a 56-43 district is relatively safe, but it’s close enough that a bit of a national shift, a strong challenger and a couple of tough votes could put it in play.
More important, for the question of House GOP behavior, a member in a 62-38 district is presumably, all else equal, going to allocate a much bigger share of paranoia to a primary defeat than to a general election defeat, at least compared with a member in a 56-43 district. To put it bluntly: if the district is 62-38, a reputation as an extremist may well be the most rational reelection strategy, but in a 56-43, district a reputation as an extremist could be dangerous.
Granted, politicians aren’t always perfectly rational in how they process their paranoia. And individual cases might be different; for example, a district that might be less Republican overall might produce tougher primaries depending on who exactly those Republicans are.
Overall, however, there’s just no question about it: partisan gerrymanders produce districts that are less safe for majority-party incumbents. That’s just math. And so we should expect those districts to be relatively more responsive to the general election than to renomination.
I do think that excessive paranoia about primary elections, and insufficient paranoia about general elections, is a significant part of current GOP dysfunction. It just doesn’t appear to have anything to do with gerrymandering.