A Republican operative who worked on congressional races this past cycle was more blunt, saying his party’s data was often wrong in contests where the Hispanic or youth voters, or both, were factors. Where a race was largely decided by the white vote, GOP data tended to be reliable, said this operative, who declined to speak on the record in order to provide a candid assessment of the surveys produced by GOP pollsters.

However, in states with small populations of ethnic minority voters, such as Montana and North Dakota, Republican pollsters surveying Senate races still misfired, which Democratic and GOP strategists have suggested was a result of failing to include the right mix of other key demographics in their turnout models, such as gender and age.

Even reliable public pollsters fared poorly in their reading of the electorate. Gallup, considered the gold standard among public pollsters, projected a Nov. 6 electorate substantially more favorable to Republicans than actually showed up.

Mark Mellman, who polled for Senate Democrats in Montana and North Dakota, said the public polls that consistently showed Republicans in the lead in those races might have been too dependent on “likely voter” screens, and thus gave a false impression to political observers that Tester and Sen.-elect Heidi Heitkamp, D-N.D., were behind when in fact they were ahead heading into Election Day. “You’re never going to have an electorate exclusively made up of likely voters,” Mellman said.