Do you remember "unskewed polls"? Do you remember all the fun America had with the (fringe, then more mainstream) conservative theories that 2012's polling was skewed to expect more Democratic voters, and underrating the coming Romney win? It was very funny, and it helped sanctify Nate Silver (this had its downsides), but it might have ruined the punter's game of challenging polls by arguing that they don't reflect the likely electorate.
Throat duly cleared, I will now ask what's up with the Quinnipiac poll of Colorado. Most surveys have shown Democratic Sen. Mark Udall, who's been running an aggressive prevent defense against Republican Rep. Cory Gardner, narrowly ahead. Quinnipiac has Gardner up 48–40, making a Udall comeback look almost impossible; out of nowhere, Gardner's soared to a +10 favorable rating while Udall is gurgling under the waves at -8. The two candidates are tied among female voters. The optimistic Republican can look at this and say that Gardner's attempt to reframe the abortion issue by promising over-the-counter access to oral contraception has driven a sword through the "war on women" meme.
But man, those crosstabs! Quinnipiac's demographic base was 80 percent white, 9 percent "other/don't know," 8 percent Latino, and 3 percent black. That's anticipating a lower Latino share of the electorate than even in the Tea Party year of 2010, when Latino voters were 12 percent of the total. In 2012, Latinos made up 14 percent of the electorate. And in both years, pollsters generally underrated the total Democratic vote in Colorado by 4 points. Quinnipiac's final 2012 poll of the state had Romney up 1; he lost by 5.
Could Quinnipiac be doing the same thing? Well, it's mid-September, and I don't know. I do know that the last Q-poll of Colorado, which had Gardner up by only 1, anticipated a 13 percent Latino electorate. Since then, the Obama administration has announced that any decision on deferring more deportations of illegal immigrants will be held until after the midterms. It is possible that Gardner's campaign has cracked the "war on women," which would surprise all the women pointing out how unsatisfying "we'll repeal Obamacare and let you spend your paycheck on the pill" is as a message. It is possible that Latino voters have taken some punditocracy advice and decided to sit out the election. Whatever the case, it's always worth checking the demographic crosstabs—unskewing be damned.