Poll Predictions and Calling a Winner (Dayton)

Continued Poll Confusion?
Poll uncertainties continue to dominate Minnesota politics as recent surveys produce vastly divergent results. The MPR poll reveals a 12-point lead for Dayton in the governor’s race, while other surveys yield a narrower lead for Dayton. How do we explain all this?

As I argued in an earlier blog, part of the problem with the surveys is that they are not necessarily capturing the underlying demographic reality of Minnesota. The MPR poll vastly overestimates the percentage of the population which is DFL and it underestimates the percentage who are independent or who have no firm partisan identification.

In addition, many of these polls are done on the cheap, failing to develop good samples that balance appropriately the percentage of the population which is Twin Cities urban, suburban, Iron Range, and greater Minnesota. Other problems are under-sampling by age, caused in part by missing cell phone-only users. There are also issues of who actually picks up a LAN line to answer a call. Some of these survey problems can be solved by increased samples sizes but that means more costly surveys.

But other problems are not solved with money. They are rooted in assumptions about what the state looks like demographically, especially from a partisan perspective. To repeat an earlier blog, here is my approximate assumption of what the state looks like:

35% DFL or leaning DFL
30% GOP or leaning GOP
10% Independence Party (IP) or leaning IP
25% Other or no party identification

Any poll that vastly departs from these approximates, such as the MPR poll, is flawed from the start. My assumptions of these percentages are based on post-2008 election exit polls and they reveal declining support for the DFL and GOP since then. The estimates are also based on talks with party leaders and analysis of voting trends, and efforts to piece together some valid data from the polls.

Finally, let me also say that my estimates are rooted in someone who, in previous a life, worked on or ran about 50 campaigns back in NY, my knowledge gained from teaching campaigns and elections and research methods (I wrote a book on the topic), my knowledge of Minnesota, and finally, some gut intuitions.

Predicting the Governor’s Race
Given the above, my prediction for the governor’s race? Here is how I analyze it.

Each of the three major candidates is doing a pretty good job holding their bases. Thus, I start with Dayton, Emmer, and Horner at 35%, 30%, and 10%. While the polls have been flawed, two initial truths come out: 1) Dayton and Emmer are holding about 80% of the bases and Dayton and Horner are capturing most of the swing or moderate voters. I am going to assume Horner is holding about 90% of his base. Emmer has been less successful in attracting the swings. Dayton is capturing about 50% of the moderates, Horner is about 30%, Emmer 20%.

Thus, here is my first calculation.

Dayton .8 X 35% = 28%. .50 X 25%= 13.75%. Dayton total= 40.5%
Emmer .8 X30% = 24%. .20 X 25%= 5% Emmer total= 29%
Horner .9 X 10%= 9% .30 X 25% =7.5% Horner total = 16.5%

However, this first calculation makes several assumptions. It assumes no undecided voters or that undecideds break in a pattern noted above. Second it assumes support for all of the three candidates is firm and that everyone will come to vote. I do not make this assumption. We know at least three additional things that correct these numbers.

First, turnout this year will be less than in 2008. In 2008 approximately 77% of the eligible population voted. Mark Ritchie estimates 60% turnout this year. I think 55-60% if realistic. This is 17-22 points less than 2008. Many of these not voting will be DFL leaning voters who came out for Obama. This means among likely voters, the smaller electorate this year will be more heavily GOP than two years ago.

Second, national polls and state polls reveal GOP voters more highly motivated than either the DFL or swings. Women (swings and part of the DFL base) are less motivated to vote compared to males. Young are less likely to vote than old (although old seem to favor Dayton). Support for Dayton is less firm, passionate, or motivated to come out and vote for him.

Third, as Election Day looms and it looks less likely that Horner can win, swings plagued by the spoiler fear (A vote for Horner is a vote for my worst or last choice) will abandon him. There is also a polarized electorate problem I have discussed. Some will opt not to vote, some wills shift to Dayton, a few to Emmer.

These three factors will produce a likely electorate more GOP than my initial numbers suggest. This is the issue of the likely voter. Correcting for these three realities is not a science, but an estimate or gut guess. I think Horner fades down to about 10-12% when a final count is done. He may even fade lower. Dayton picks up some Horner votes and the more GOP leaning electorate yields this as my estimate.

The Winner is . . .Dayton!
Dayton 46%
Emmer 43%
Horner 10%
Other 01%

I am guessing Dayton has about a 3% lead in the polls, it may be as high as 5%, depending on whether my initial guesses about the size of the bases, capture of the swings, and shape of the electorate are accurate.

I am not offering science here, but best estimates or guesses. Now let the voters decide.