Why the 20-point spread in KSTP, Strib Minnesota polls? They’re BOTH outliers

As if to underscore what an erratic year it’s been in political polling, the latest Minnesota numbers in the presidential and US Senate races come sporting a gap like none I can remember in any past contest here. Last week SurveyUSA and KSTP released numbers indicating adead heat between John McCain (47 percent) and Barack Obama (46) and a 10-point lead for Norm Coleman (43 percent) in his race with Al Franken (33) and Dean Barkley (19). Over the weekend, the Star Tribune/Princeton Survey Research Associates poll produced an almost diametrically opposed portrait of those races:Obama 55-37 over McCain, Franken (43) up by 9 over Coleman (34) and 25 over Barkley (18). SUSA/KSTP was conducted on 9/30 and 10/1; PSRA/Strib on those two days plus 10/2.

What the two surveys lack in coherence when placed together, they make up in symmetry: In both races measured, SUSA and PSRA are exactly 19 points apart in their margin spreads (McCain +1/Obama +18, Coleman +10/ Franken +9). Typically when polls disagree by a substantial margin, a spat ensues over which one is the outlier. But when polls are this far apart, there’s a common-sense presumption in favor of thinking they’re both outliers.

But it’s hard to be much more specific, because neither poll offers full questionnaires or demographic cross-tabs. It’s tempting to think that the SUSA/KSTP numbers reflect an oversampling of Christian conservatives (in other words, outstate Minnesota). It’s a large and generally underrated population in Minnesota, after all, and precisely where the McCain campaign has gained the most traction nationally since adding Sarah Palin to the ticket–and SUSA’s month-to-month numbers show Obama falling by 3 points in Minnesota versus McCain since the last, pre-Palin state survey in September.

SUSA’s own write-up of its poll notes an unprecedented, and unexplained, new element of Minnesota exceptionalism in its results: “Minnesota behaves unlike other states in some respects. Among women, there is movement to McCain, at a time when McCain is losing ground among women elsewhere. Among voters younger than Obama, there [is] movement to McCain, at a time when Obama is consolidating support among young voters elsewhere. Among voters older than McCain, there is movement to Obama, at a time when older voters elsewhere are sticking by McCain.” Can you explain any of this in light of observable recent trends? I can’t.

In similar fashion, the PSRA/Strib numbers seem to augur for an oversampling of Democrats (in other words, the metro area). Key passage: “Forty-two percent of respondents identified themselves as Democrats in the latest poll, up from 34 percent a month ago. In the same period, the percentage of Minnesotans calling themselves Republicans has dropped from 31 percent to 26 percent.” In one month’s time, after these races have spent more than six months in the public eye? Why? This seems as unexplained, and nearly as suspect, as the SUSA numbers saying McCain and the Republicans have actually gained ground in MInnesota in the past month.

Since there is limited detail on both these polls, how do they measure up against the most recent polls by other outfits assessing the same races?

The most recent Minnesota presidential survey aside from these two is a CNN/Time poll conducted from 9/28-30, a couple of days prior: Obama 54, McCain 43.

The most recent Senate poll is an internal survey by the Mellman group for the Democratic Senate Campaign Committee–obviously a partisan organization, but one with an interest in getting accurate numbers since their decisions about where to throw campaign money are predicated on the results: Franken 38, Coleman 36. (Before that, a trio of mid-September polls had given Coleman a lead ranging from 1 to 7 points.) For what it’s worth, these comparisons to other recent polls suggest the Strib poll is closer to the consensus view of pollsters than KSTP. But in a season where survey-takers everywhere are confounded about what a “likely voter” looks like, and a growing portion of the public–especially under-30s–aren’t even reachable by standard polling practices, we’ll leave it to you to decide how much comfort or distress to find in that fact.