Political scientists Jowei Chen and Jonathan Rodden, writing in The New York Times, argue that Republicans won their majority in the House of Representatives in 2012 not so much because of favorable redistricting as from the clustering of Democratic voters. They developed an algorithm that allowed them to construct thousands of congressional redistricting plans for each relevant state (states are considered irrelevant if they have only one congressional district). They found that Republicans gained some seats over a random redistricting in Florida, Michigan, North Carolina, Pennsylvania and Texas, but that Democrats also did so in Illinois, Maryland and (by gaming a supposedly nonpartisan commission) California.
This backs up a point I have been making for some time: Heavily Democratic constituencies -- blacks, Hispanics (in most states) and gentry liberals -- tend to be clustered in central cities, certain sympathetic suburbs and university towns, which gives Democrats enormous percentages. But Republicans are spread more evenly around the rest of the country.
In 2012, some 80 congressional districts gave 70 percent or more of their votes to a presidential candidate -- 61 to Barack Obama, 19 to Mitt Romney. Those districts performed as you might expect in House races: 61 Democrats and 19 Republicans. The rest of the country's 355 congressional districts, which voted less than 70 percent for a presidential candidate, elected 215 Republicans and 140 Democrats. Redistricting played some small part in that, but the main culprit -- or hero, depending on your point of view -- is clustering, as Chen and Rodden conclude.