Gaming The Presidential Polls

Most Americans have heard the weekly and sometimes daily voter polls letting us know there is no need to vote because Obama has it sewed up.  But few  treally understand how  statistics can be “gamed” to fit an agenda.

The recent presidential poll in Florida found Obama was leading Romney 54 to 44 percent in the state.  However,  when you look at the fine print, the breakdown of poll numbers is quite different.  Obama leads Romney 84 to 5 percent  among Democrats; Romney leads Obama 91 to 7 percent among Republicans; and Romney leads Obama 49 to 46 percent among Independents.  Obviously there must have been at least 10% more Democrats surveyed  than Republicans to get Obama up by 9%.

When you question the validity of this poll,  defenders will say they are using 2008 national numbers of Republicans, Democrats and Independents.  However, in 2008 the party breakdown in Florida was 38% Democrat, 34% Republican and 29% Independent.

Assuming Obama is as popular as he was in 2008, which every major poll shows is untrue, and using the 2008 numbers of Democrats, Republicans and Independents in Florida, the poll would show Obama leading Romney 50.5% to 47%, which is a 3.5% differential, much different than the 9% reported.

And, if the proportion in the state is 37% Democrat, 36% Republican and 27% Independent, which is more likely, the poll would show Obama still ahead, but only by a margin of 49.7% to 47.8%, which is 1.9%, not 9%.

The myth of pinpoint accuracy in election polls is a myth.   Any Poll is an estimate!

It is hard for voters to have reliable evidence as to whether a poll is accurate, close to accurate  or deliberately skewed in one party’s favor.  Besides the professional pollsters like Gallup or Rasmussen, there are partisan polls and others that conduct private candidate surveys.

Polls are conducted in a number of ways – some call only registered voters, some call cellphones and some don’t; some use computerized voice recognition systems, some rely on a small sampling to determine a nationwide poll, etc.

Once all the raw data is assembled pollsters apply various mathematical “weightings” to their finding.  That is because almost every poll, no matter how many people are polled, has some weakness.    A common problem is finding enough young males to respond so pollsters extrapolate from the number of young men who were reached to a larger sample of young males assuming that the small sample will reflect the  larger sample.

Polls can be skewed by the breakdown of Democrats, Republicans and Independents.   Some pollsters believe there is no way to know exactly who will turn out to vote on election day so they assume if they use a truly random sampling, the breakdown of Democrats, Republicans and Independents will take care of itself.  Others try to forecast voter turnout based on recent surveys that screen for likely voters.

How questions are asked can also have an impact on the answers pollsters get.  For example, if you ask whether they support or oppose the death penalty for  convicted murders, about 69% of those polled will support it.  But, if you ask if they prefer the death penalty or life in prison without parole, the percentage of support for the death penalty drops in the 40s.

Then you need to look at the “margin of error” which is a percentage based on the size of the poll.  Take last weeks Pew Poll – they had Obama leading Romney 51-43.  Rasmussen’s poll came out the same day and they had Romney leading Obama 47-46.    While it may look rather strange to have two polls with such opposite results, when you apply the “margin of error” to both they are pretty close.

The Margin of Error in the Pew poll was 2.4% which can be applied to both numbers.  In the best case scenario for Obama, the Pew poll could show Obama leading 53.4 to 40.6  (add 2.4 to Obama and subtract 2.4 from Romney).  At the same time, the best case scenario for Romney would have Obama only leading 48.6 to 45.4 (deduct 2.4 from Obama and add to Romney).

In the Rasmussen poll there was a 3% margin of error so in the best case scenario for Romney he would lead Obama by 50-43 or Obama could be leading Romney 49 to 44.

According to Tim Vercellotti, a political scientist and director of the Polling Institute at Western New England University you “can’t look at them [polls] as any kind of guarantee that this is how the world really is.”   Gary Langer, president of Langer Research Associates who polls for ABC, says that  “Unfortunately there’s a lot of ways to manufacture data.  It happens all the time.”




Print Friendly, PDF & Email

Leave a Reply

Your email address will not be published. Required fields are marked *