Political Hay

## Polling for Yourself

How can John Kerry win if most people tell you it wonâ€™t be easy?

9.17.04

Send to Kindle

About a month ago, I began mailing around what I called a "personal psychographic" question to various friends and Internet acquaintances. The question aimed to tap a person's inner knowledge in order to effect a prediction about the election. Rules were simple: You assigned a percentage to each of four categories. All four categories had to add up to 100%.

The four categories were:

Kerry landslide
Kerry close win
Bush landslide
Bush close win

Play around with these categories a while, and you'll discover what's so revealing about them. Even a Kerry partisan would -- I'd guess -- assign no more than a 5% chance of a Kerry landslide. A more realistic guess would be zero. Let's say 2%, just to be generous.

A close win for Kerry? Give the Kerry partisan his innings again, let him (or her) consider that he's already assigned a 5% probability to a Kerry landslide, and it'd be hard to figure odds much higher than 50% of that happening.

Okay, so your Kerry partisan is comfy now with a 55% overall prediction of a win.

But wait. That partisan still has to predict (just for him or herself, not for anybody else) what the likelihood of the Bush margins are. And I think even a Kerry partisan might grudgingly admit a 7-10% chance of a Bush landslide. Let's say 7. And, still grudging, let's concede that Bush might have a 45% chance of a close victory.

But wait. Remember the rule: All percentages have to add up to 100. If Bush's total grudging percentages are 52, what are we to do with Kerry's purported 55? From whom do we subtract what?

Fool with it yourself, regardless of your sentiments. Try to be as realistic as you can. You will find it very, very hard to predict a Kerry win.

Our own esteemed editor, Wlady Pleszczynski, summarized a similar poll last week in his "Editor's Desk" entry this way:

…America Online polled its users. Normally AOL's questions lean N.Y. Times left. But this time there was no beating around the bush. The replies from 322,440 plus respondents (at last count) couldn't have been blunter: 41% predicted Bush wins easily; 29% said "Bush wins tight race." By contrast, only 26% said "Kerry wins tight race," and a pitiful 4% noted "Kerry wins easily." Any way you look at it, that spells Bush over Kerry, by 70 to 30.

Wlady rightly points to the predictive value of that kind of poll. And a kind of inverse (and perhaps perverse) note about poll reporting this election cycle also points to a Bush win. Virtually every poll taken includes such a "Who do you think" question. This year, those results have almost never been reported. Inverse conclusion? The mainstream press doesn't want you to know that, all along, most people have thought Bush was going to win.

Lately, however, the news has broken through.

The September 14 IBD/TIPP poll from Investor's Business Daily, while headlining a 47-47 tie between Bush and Kerry in overall results, concluded its report: "The zeitgeist still points to a Bush victory. Some 48% of those polled foresee Bush retaking the White House, while only 16% think Kerry will win. Another third (31%) feel the race is too close to call."

Grand Rapids, Michigan, TV station WZZM published the results of a "Who Do You Think Will Win?" survey the next day. Their results? Bush 59%, Kerry 37%, Someone Else 2%, Not sure 2%.

Regional polls have carried the same news. In Iowa, as reported on MSNBC.com September 2, a Survey USA poll found that "More Iowans Think Bush Will Win" -- meaning the state, not the nation. Some 60% of the people surveyed thought Bush would carry Iowa. (The report gave no further data.)

While these two polling methods -- mine and the press's -- have some things in common, they are, in fact, very different. The media polls strictly measure horserace sentiment. They are valuable predictors of a winner, but not of a margin. Surely, nobody expects Bush to win 70% of the vote.

But, with a large enough sample of my internal psychographic poll, I think one might actually predict a margin of victory. Suppose you averaged the results from every respondent for every category. You would probably come up with something like this:

Kerry landslide 1%
Kerry close win 46%
Bush landslide 7%
Bush close win 46%

Predicted margin of victory: Bush by 6%. Adjust for Nader, uncertainty, etc., and I'd say you come up with a Bush victory by 4-4.5%.

How about it Zogby? Rasmussen? Gallup? Any takers on this methodology?