The most recent round of Quinnipiac polls re-ignited Democratic fears that Donald Trump could actually win the presidency. The polls put him ahead of Hillary Clinton by 4 in Ohio, and behind her by only 1 point in Florida and Pennsylvania.
Democrats should not be complacent about the threat of Trump, but neither should they revert to panic at the sight of a single poll. Always beware of outliers.
First, if you average all of the polls in each state taken since March, Clinton holds a 5.6 percent lead in Florida, a 7.0 percent lead in Pennsylvania and a 3.0 percent lead in Ohio. None of states are a lock for Clinton, but the Quinnipiac results are a clear outlier.
(For the record, Bernie Sanders' average lead over Trump in those polls since March is only significantly better in Pennsylvania: 5.0 in FL, 5.0 in OH, 12.7 in PA.)
Why is Quinnipiac an outlier? One reason is it assumes the electorates in those three states will be more white than in 2012, when a joint effort by the Center for American Progress and the American Enterprise Institute project, dubbed States of Change, projects each state will become slightly less white in 2016.
The other odd aspect of the Quinnipiac poll is that it does not show Trump performing better with whites that Mitt Romney, but it show Trump performing better than Romney with people of color.
In Florida, Quinnipiac finds Clinton beats Trump among nonwhite voters by 43 points. But Obama beat Romney by 49 points.
In Ohio, Clinton's margin is 62 points. For Obama, it was 79 points.
In Pennsylvania, Clinton's margin is 60 points. For Obama, it was 85 points.
So to believe that Quinnipiac is right and the other recent polls are wrong, you have to believe that the electorates will be whiter than 2012 and Trump will win more people of color voters than Mitt Romney in 2012.
Most importantly, polls are not votes. The electorate will be shaped not by pollsters but by the efforts of the campaigns and their volunteers. If you are still nervous that the Quinnipiac assumptions could prove to be true, then go knock on doors, turn out the vote and prove the assumptions wrong.