Poll Analysis: Paul Ryan Pick a Big Winner

The selection of Paul Ryan to be Mitt Romney’s running mate has electrified conservatives and appears to have boosted Romney both nationally and in key battleground states. While Nate Silver of Real Clear Politics has estimated that the Ryan pick has been worth only two points so far to Romney, I believe it currently could be worth more than that.

It is likely too early to measure the full Ryan impact. The Gallup poll is a seven-day sample; some of the days in its latest daily average include pre-Ryan selection days. There are not yet any national surveys by the once-a-month pollsters conducted entirely after Ryan was named.

The two tracking polls are in the market every night with 500 interviews each; both Gallup and Rasmussen have shown a several-point swing to Romney in the last 10-14 days. Rasmussen has moved from Obama +2 to Romney +1. Gallup has moved from Obama +4 to Romney +2. Because Gallup is a seven-day sample, its daily average generally changes less rapidly than Rasmussen.

Both polls have been fairly stable for the last few months, with Romney's and Obama’s numbers in the 45-50% range, so the shift in both surveys is likely not happenstance. Some of the shift occurred after the Ryan pick and some before --possibly blowback from the “Romney caused my wife to die of cancer" ad by the Obama Super PAC.

Nate Silver’s model produces a two- to three-point lead for Obama at the moment. The tracking polls suggest otherwise.

For a few weeks, a series of national polls by once-a-month pollsters -- Fox, Reuters, Pew, IBD, and CNN -- suggested that Obama was pulling away. The Obama lead was 7-10 points in each survey. But each of these polls included samples with far more Democrats than Republicans. Given that so far this year most every survey shows Democrats giving 90% support for Obama and Republicans giving 90% support for Romney, oversampling one side or the other by five or ten points turns out to be the entire story in a poll. Some pollsters, such as Rasmussen, attempt to keep the party shares among those interviewed each night roughly consistent with shares for the two parties and for independents, based on support for candidates in prior elections and other measures.

If you get a sample in a survey that tells the pollster it supported Obama by 14% in 2008 when he won nationally by 7%, or that it supported Obama by 13% in Florida when he won the state by 3%, you likely have too many Democrats or Democratic-leaning independents in that sample. Silver has admitted that some polls -- such as Pew, Reuters, and PPP -- have had large slants towards the Democrats this year (three to four points).