Saturday, October 25, 2008

The Two-Track Hypothesis of Voter Decisioning

Earlier this campaign season, I began to question the polling methodology being used, especially when compared to historical norms. Polls released at the same time, claiming to use the same methodology, were publishing results well outside the range of their margins of error, demonstrating fundamental mistakes in their models. While some of those polls played fast and loose with racial, economic, and age demographics, and at least one major poll grossly over-sampled unemployed adults and another poll which had published its demographic internal data through 2006, for this election stopped revealing that data, the most common disparity between poll models has been political party affiliation. These have come in two flavors – polling groups which have oversampled democrats in the belief that democrats will dominate the actual voting to a degree not seen in most Americans’ lifetimes, and polling groups which do not weight their samples for party affiliation but merely report the proportion of party affiliation of the people contacted. The first assumption is based completely on subjective prejudice and in some states is wildly variant from the actual election support from the 2006 (last federal) and 2004 (last presidential) elections. The second assumption is absurd on its face. To illustrate, I could have taken a poll at the Democratic and Republican conventions this year, and covered pretty much all of the census demographics, including gender, age, education, work background, geographic hometown, urban/suburban/rural split, religion, and so on. Yet I think we can safely say that polling only democrats or only republicans would produce a poll which would be absolutely useless in telling us how the nation really felt; political affiliation is undeniably a significant vector in voter support for a candidate. You do not have to be an expert in political analysis, to understand that democrats and republicans will overwhelmingly support their party’s nominee for president, and so increasing the proportions to favor one party in representation will unavoidably skew the results in favor of that party’s nominee.

I have noted before that history shows a remarkably stable proportion of party affiliation, the democrats generally outnumbering the republicans by between 2 and 4 percent. A poll, therefore, which assigns 10 to 15 percent higher participation nationally by democrats in a presidential election is simply unsupported by any historical sample in decades. This raises a valid question, though: Why then are so many people taking part in polls calling themselves democrats? The answer to this question is important to understanding not only why I believe the polls are wrong for the most part, and why the election strategies of Barack Obama and John McCain have always been different by need as well as design.

If you look at the kinds of arguments between democrats and republicans, especially between liberals and conservatives, you may note that the dialogue generally breaks down early. This is not only because common ground is so hard to find, but because the motivations are different. Someone rang a bell in my head earlier this season, when they noted that Obama supporters generally support him because of how he makes them feel. The person I was speaking with, was explaining this as one reason why Hillary Clinton did not win the primaries early on; she did not make democrats feel excited the way that Barack Obama did. I have also noted that people who are still undecided, frequently say that they have not yet made up their minds, that they have questions for which they want answers from the candidate they are considering. In a nutshell, these are the two types of decisioning with voters; some make their decision largely on emotion, while others make their decision largely on intellect. That’s not a democrat/republican thing all the time, nor is it that one type produces winners more often than the other, and it’s not that people are one or the other; I believe we all react both emotionally and intellectually for or against a candidate, and our personal balance makes the decision. But it does explain how support is collected for a candidate, as the emotional commitment is made far earlier than the intellectual buy-in; in fact I suspect that almost all last-minute deciders are heavily influenced by intellect rather than emotion. If a candidate is charismatic he can win over the emotional base, but an experienced candidate is more likely to claim the intellectual base. When a candidate is able to address both types well (a Reagan or an FDR, for example) then you see landslides. If a candidate is grossly unqualified in one of those venues, then he may lose in a landslide (like McGovern or Goldwater).

If this theory is correct, then obviously Barack Obama has the advantage in emotion-based campaigning, while John McCain has the advantage in intellect-based campaigning. Obama’s lack of experience makes it very difficult to build a case for him on accomplishments; he simply has no resume. McCain’s lack of glamour makes it very hard for him to gain, much less hold, the attention of anyone not already inclined to give him a chance to make his case; he simply does not sparkle. The question at hand, however, is which approach is more effective in this year’s campaign. The polls would seem to indicate that Obama grabbed most voters’ attention, won them over, and they never gave McCain a serious look. That, however, ignores an obvious side-effect of the emotion-based voter. Pollsters this year have – among themselves – remarked about the difficulty in getting responses from people they contact. Some of this is blamed on new technology and the fact that many people spend less time at home to be contacted, but it is also an important historical fact that democrats have traditionally always been more interested in taking part in polls than republicans, and this year the emotion-based voter is much more inclined to take part in a poll to discuss how he feels, than an intellect-based voter who wants to make sure of his vote before he tells anyone else about it, and who in any case has no particular interest in talking to a stranger about how he feels. As a result, the polls may be feeding off their own assumptions, using the circular logic that the results from their skewed polls justify the bias. If I am correct, more than a few polling groups will be doing a lot of work come December and January to try to figure out what went wrong. The fact that the practices at these groups do not include applying a Deming loop, is a warning sign they missed years ago, so I am skeptical about their ability to learn. The worst-case scenario from my point-of-view, is that the effect of these invalid polls might dismay republicans enough to stay home and not vote, creating the sort of disparity in voting patterns to indicate the polls were right, so that they might never consider that their bias could be creating the effect. We shall see. Obviously, I have to admit that I could be wrong, so in the event that Obama wins all 60 of the states he calls America (where he gets the additional ten states, he has never made clear), I will be reviewing my own work in the interest of honesty and that same Deming loop I was just talking about.

I would like to make a few points in closing this article. If I am wrong and Obama is really crushing McCain, there are certain indicators which will show this. First, the early voting should be much, much heavier among democrats than republicans, and the youth vote we heard so much about should be a big part of the early voting. Second, we should start to see pan-demographic support in the polls for Obama in all geographic and age groups, since this happened in Reagan’s 1984, Nixon’s 1972, and Johnson’s 1964 landslide victories. And third, since Obama enjoys strong support in heavy-population states like New York and California, if he is going to collect 340 or more electoral votes, we should see evidence indicating he will reach 57 or 58 percent popular support nationally. Conversely, strong republican turnout in early voting is an indicator of stronger McCain support than has been indicated. If McCain continues to hold support in the same demographic groups he held in mid-September, again this indicates a much closer race, and since so many of the “red” states are less dense in population, any indication of a close popular race would support expectation of a close electoral race as well, since McCain could strategically win the electoral race as Bush did in 2000, with less popular support nationally than Obama but winning the necessary electoral votes.


Hestrold said...

DJ - the wrap-up after the election will be incredible. As a dispirited McCain supporter the constant flood of reverence for Obama and the drumbeat of defeat for McCain on the part of the media, the pollsters, and the Democrats leaves me with no idea where things stand. Everyone else seems to thing it's a landslide of epic proportions. Yet I go back to what this man stands for, his associations, his beliefs and I can't get my arms around my America, that would even consider such a turn from the freedoms that we hold so dear.

jwill46 said...

Regarding the early voter turnout, Hillbuzz has a story about the first set of early voting figures from Colorado:

The state of Colorado has released its first set of early voting figures: and it’s bad news for Obama.

Despite a massive push by Obama for early voting, including repetitious pleas during the Democrats’ Convention for Coloradoans to early vote, Democrats and Republicans are tied in terms of early voting so far.

105,277 - Democrats = 49.92%
105,615 - Republicans = 50.08% √
338 = Difference between the two = 0.16%

This means that even with a nonstop push by the Obama campaign for his followers to vote early, Republicans are still ahead in Colorado, if only slightly.

Anonymous said...

Was reading an article on party identification in the military, and came across this interesting point:

Quote: "According to data collected in 2004 by the Annenberg Public Policy Center, 37 percent of the civilian, non-veteran population identified as conservative, 24 percent as liberal, and the remaining 39 percent as moderate or undecided."

Annenberg Public Policy Center is a credible org, and despite the shift over the past 4 years toward the Democrats, I cannot imagine how polling orgs think the Democrats have gained 15 points over the Republicans in that 4 years. I think your arguments on weighting and other questionable polling devices have merit. Plus, as a Democrat who voted for McCain/Palin, I think pollsters may be missing the 18-20% of former Clinton supporters who reject Obama as a legitimate nominee. Of course, we'll see on November 5 [hopefully it's not a protracted court battle again] whether or not the weighting reveals itself to be a major flaw, in which case, pollsters have committed professional suicide, in my opinion. Either way, a dark national depression is going to set in about February/March, 2009, when America realizes that no matter who's in the Oval Office, we're screwed. The economic crisis won't go away anytime soon; and it will get much worse before it gets better. Even when it improves, there will be fundamental changes to this society which will impact the next generation in ways they never would have dreamed. Cheers!

Lea said...

I think your argument makes a lot of sense. Is it plausible that the pollsters are purposely trying to impact the race, purposely trying to make Republicans think we are losing so badly we should just stay home?

Hestrold said...

Lea - I don't think the pollsters blatantly use their bias to cook the books on their polls. Since they are human and arrogant at the same time they are blind to their own bias. At the same time there are so many shifting data points that they truly are having a hard time coming up with a representative sample. If they were measuring something that lent itself to a scientific analysis the polls would be more consistent. Since they are not, that means it's more art than science. So who knows what will happen? I don't. And neither do the pollsters.