The Schultz Report audiocast is on hold while David Schultz is traveling in Europe and parts of the former Soviet Union on a speaking tour about the U.S. elections. But we managed to catch up with Schultz via e-mail earlier today in Finland (yesterday it was Estonia), where he graciously took time to answer a couple of questions about races back in Minnesota — specifically, what he made of the diametrically opposed Minnesota Senate and presidential polls we wrote about Monday, and why Senate candidate Dean Barkley seems to be taking a disproportionate amount of his support from Norm Coleman.

Minnesota Independent: As you’ve probably seen, a pair of hugely discrepant polls came out last week and over the weekend from SurveyUSA/KSTP and Princeton Survey Research/Strib. (numbers are broken out, and original stories linked, here: Can you shed any light on how two surveys that purport to measure the same thing can come up so far apart?

David Schultz: As someone who has taught statistics and research methods and who wrote a book on the subject, I know something about doing polling and surveys.

The fundamental problem in reconciling which of the recent two polls by KSTP and the Star Tribune is more accurate resides in knowing two things that we do not know.  First, we need to establish a correct baseline for what the Minnesota electorate demographically looks like. Second, we need to know far more about the definition of “likely voter” or the sample used to make the statistical claims in these surveys.  Unfortunately, we have neither.

What we do not know for sure is what percent of the population in Minnesota now considers itself to be Democrat, Republican, or Leaning towards one of these two parties.  If we knew that, then we might be able to make some good projections regarding whether we have over or under sampling of specific populations.  This would tell us something about survey bias and validity. The two polls reach very different numbers about how many people identify themselves as a member of one or another party.

One way to determine which poll is more accurate is to look to the past for some data. In June, 2006, we had survey results that suggested 36 percent of the MN population identified itself as Democrat or Leaning Democrat, with 35 percent saying Republican or Leaning Republican.  After the 2006 election and up through 2007 some evidence suggested that the numbers had shifted to about 41-32 percent for Democrat/Leaning Democrat and Republican/Leaning Republican. These numbers are similar to national party ID trends, although in recent weeks the numbers are closer to about 40-33 percent.  This is a seven point party ID spread.

However, this number does not tell us the entire story for establishing a baseline.  We also need to know more about who the surveys sampled and considered to be a likely voter.  In doing surveys, the most crude way to determine if one is a likely voter is to ask.  This is a bad test because people often lie about whether they have voted or will vote.  A second way to ascertain who’s a likely voter is to ask other questions.  For example, one could ask if one voted in the past.  Past performance is often a good predictor of future behavior.  However, this does not work with first time voters, especially those who are young.  Their failure to vote in the past may not tell us anything about current plans to vote.  This might be especially true this year when Obama has targeted the youth vote and it is highly motivated.

We should also remember that about 15 percent of those who vote in Minnesota register on the day of the election.  This group is under-reported in polls.  Of course it is almost cliche to point out that many young and middle age people either do not answer landlines for surveys or do not have landlines but instead use cell phones.  Cell phones are often not used when calling for surveys.  All of these people being missed are probably Democratic Party leaning.  I argue this because other surveys of political attitudes of young people show this, and because many studies of voter registration across the nation show tremendous Democratic party enrollments this year.

The point I am making is that there is good evidence that a 7-point party spread is conservative and that perhaps a spread of maybe 10 points makes sense.  If we accept that the Democratic/Republican spread is about 42-32 percent, that gets one closer to the Star Tribune poll but still not the 16 point spread that poll suggests.  This 42-32 percent is far closer to my estimates.

Given these numbers, I doubt McCain and Obama are tied, but I also doubt an 18-point lead.  My estimate would be 10-12 point lead, perhaps a little more.  We need to factor in the points about the youth vote and party ID noted above against the very strong tendency of religious conservatives to show up and vote.  My estimate, the recent CNN poll is closest to what is perhaps reality.

MnIndy: In the Minnesota Senate race, it appears by most measures–except the SUSA poll–that Dean Barkley is continuing to draw support mainly at the expense of Norm Coleman, while Al Franken’s numbers remain stable in the low 40s. What’s going on here, in your view? It almost seems as if there was a larger-than-realized pent-up demand for almost anybody but norm, but that many of those folks would not go for Franken. Do you think that’s so, or have the vitriolic ads that Coleman has run this fall really hurt him that badly?

Schultz: Norm Coleman is the accidental senator to many.  But for a plane crash six years ago he is some flunkey in Bush administration.  He has never been overly popular and Democrats are highly motivated to beat him if they can find the right candidate.

It is unsure if Franken is that candidate. He pretty much had held close to the overall base of Democratic/Leaning Democrat voters (see above) the entire election season, with about 40-42% of the vote.  His numbers have not changed much at all and I suspect that he is mostly getting a lot of his base (not all as the DFL primary shows) and a few swing voters who cannot stand Coleman.

My suspicion is that when it was a two person race more swings disliked Franken than Coleman and it therefore worked to the latter’s advantage in going negative.  However, Barkley entering the race changes everything.

A few years ago I edited a book on political advertising.  One chapter by one of the authors pointed out that negative ad wars usually depress voter turnout by 3-5 percent.  It does that by disgusting mostly swing voters who opt not to vote for any one.  Yet the depression of voter turnout thesis might not work in a three way race.  Here instead of swing voters saying “I’m not voting,” they might say “The hell with the two of you, I am voting for the third party candidate.”  Here the beneficiary is Dean. Individuals who do not like Coleman but who cannot stand Franken now have an option, they vote for the third party candidate.  Thus, while the attack wars have whittled Franken and perhaps Coleman down to their bases, the rest of voting for Barkley.

Franken can win with a perfect storm.  If he keeps going negative he drives more voters to Barkley and away from Coleman.  Now assume more Minnesotans are Democrats, that there will be 80% turnout this year as the Secretary of State predicts, and that there is an Obama coattail effect.  Put all together and Franken wins with 40% of the vote.

Overall, using a baseline of DFL/GOP ID of what I suggested above and given the assumptions made here, Obama is probably winning by 10-12 points and Franken and Coleman and roughly tied.

What to look for?  Will McCain pull a Michigan in Minnesota soon and cut his money here?  If yes that is a clear sign of the polls.  But after Michigan McCain may not be able to afford giving up on Minnesota, even for symbolic reasons.  I never thought MN was winnable for McCain–with out without Governor Pawlenty on the ticket–and I am still unpersuaded it is a GOP pickup this year.