Voter Polls - Evaluating The Numbers

---

ELECTION 2006 / VOTER POLLS: EVALUATING THE NUMBERS & PERCENTAGES

---

Don't Look At Polls, Look At Trends: Because of the margin of error built into every poll and the occasional statistical blips that cause polls to blow out in one direction or another, any one poll isn't necessarily all that meaningful (although of course, you'd always prefer to be ahead, rather than behind, and the bigger the margin, the more chance that it accurately portrays who's ahead and who's behind.)

A Simple, Three Step Guide To Getting The Most Out Of Election Polls

~ John Hawkins

You hear a lot of people saying that you can't trust the polls and that they're always slanted against the GOP. That's actually not true. Rather than go into a long, detailed breakdown on polling data, let me give you some short and simple tips on how to actually get meaning out of the poll numbers you're going to see in abundance in the run-up to the election.

• #1) What Type Of Voters Are Being Polled: Polls of adults? They're meaningless for political purposes and slant heavily towards the Democrats. Polls of registered voters? They're a bit more meaningful, but they still slant towards the Democrats. Polls of likely voters? These are the most meaningful and accurate polling numbers for election results.

• #2) All Pollsters Are Not Created Equal: In my book, there are two types of elections pollsters.

The first is reputable pollsters including, but not limited to: Gallup, Rasmussen, Survey USA, Quinnipiac, Mason Dixon, etc. The pollsters aren't perfect, but you can have some confidence in their numbers.

The second group of pollsters includes universities and newspapers in general, internal polls, polls done by partisan polling companies, and Zogby. These polls aren't meaningless, but you should really take the numbers they put out with a grain of salt.

• #3) Don't Look At Polls, Look At Trends: Because of the margin of error built into every poll and the occasional statistical blips that cause polls to blow out in one direction or another, any one poll isn't necessarily all that meaningful (although of course, you'd always prefer to be ahead, rather than behind, and the bigger the margin, the more chance that it accurately portrays who's ahead and who's behind.)

So, in order to get the most value out of polling data, you need multiple polls from multiple polling agencies to get a real sense of who's ahead or behind.

For example, show me three polls from a month ago showing a candidate down by an average of 15 points and three polls from last week showing the same candidate down by an average of 4 points and you can comfortably predict that candidate has momentum.

Show me three polls, all from roughly the same time period, showing a candidate down by 4, 1, and 5 points, and you can feel comfortable in saying that candidate is really down.

Show me three polls, showing a tie, one candidate up by 2, and the other showing the other candidate up by 3, and you can comfortably predict that the race is within the margin of error.

Now, will this always work? No, because when you play with statistics, you're going to come up with some oddball numbers every once in a while. But, it will usually work and if you follow trends, not individual polls, you'll have a pretty good idea of what's going on as long as there is enough polling data to work with.

© Copyright 2001-2006 John Hawkins

Send A Link For This Article To A Friend Send an e-mail message with a link to this article to anyone/everyone in your address book. Click on e-mail [envelope] icon, below

---