BLAMING THE POLLS

WHY LOSING CAMPAIGNS BLAME THE POLLS

Kenneth Hacker

November 2, 2008

Today, Karl Rove let the cat out of the bag while being interviewed by FOX news host Chris Wallace. Wallace asked for a bottom line election likelihood read and Rove said that the finishing days are rough for McCain to win. Obama is ahead no matter what the margins and he is most likely to keep an electoral college vote lead. However, there is always a chance of losing for Obama, but Rove then launched into a statement about how polling announcements can affect voters and bad news can depress voting turnout for the losing candidate. Yes, indeed. He is right all the way around.

When campaigns are winning, Rove knows, and everyone in the business will tell you, they talk about being ahead in the polls and gaining support of the American people. When campaigns are losing, they will tell you that the only poll that counts is the one done on election day. They may also throw in some barbs about pollster bias, invalidity, etc. When really uninformed, they may talk about Dewey vs. Truman when sampling was not yet perfected or employed and Dewey was announced as the election day winner as the presses rolled on election day eve in 1948.

I heard one poor soul on the radio this weekend opining about how we should be not trust polls for predicting anything since they are only snapshots of voting opinion. What this person does not realize is that a series of snapshts when done well can give you some good ideas about what follows in sequence.

Can you stick the fork in McCain now? Maybe, but there is a chance that he can win despite the statistical likelihood of him losing. This is not due to polling error as much as it is the volatility of undecided voters or weakly decided voters.

Samping error is very real and partially accounts for various polls being discrepant. Still, the smart consumer is not relying one one poll at one point in time. What gives you prediction ability is an analysis of multiple polls over a course of time accounting for margins of error and possible sampling problems. Regardless of how precise polling data, there is always a possibility of error in the estimates. On the other hand, historical patterns indicate that good polling, such as that done by Gallup, is reasonably close to what occurs on election day and what we see at this stage of the game has predicted election day winners for decades now. Upsets are very rare and occur only when major events move things in the upsetter's favor. These are historical events like the Iran hostage situation which contributed to Reagan's upset against Carter in 1980.

Likely voters data is more important that data just about registered voters because the former are more likely to vote than the former. Even better are data about those who have voted in the past and intend to vote now. Gallup pollsters use the latter in their traditional model of voters. The model using intent to vote now is called the expanded model. The traditional model, again examine over many days, can give you some prediction power regarding who is most likely to win.

Today, Gallup, in its traditional model, shows Obama at 53% and McCain at 42%. Using intervals for margins of error to see if McCain has a chance at this late date shows the following. With a 2% margin of error, we that the best case for McCain is having a 44% while Obama has a 51%. This means that his campaign is finished. Obama's best case is a 55% to McCain's 42%.

John McCain is to be honored for his military service and his service in the United States Senate where he was known for working with bipartisan planning and negotiating. His campaign is not the failure of his own stature as much as it is the failure of his consultants and his running mate who turned his campaign into a circus and a nightmare of bad strategies.

Obama, now poised to become the 44th President of the United States, has an incredible challenge of helping American to heal from a bloody election and from its past mistakes it is always growing away from.