Not All Polls Are Created Equal

Daily, we see different polling companies issue their snapshot of where the fall election stands at this precise moment. One would think that if all the pollsters did their job correctly that their polls would be in sync, but as of a few days ago some have the election at a dead heat, and others had Barack Obama up by at least 4 points. To understand how polls can be different, let's dissect the methodology of two specific ones:

First, for example, the recent Rassmussen poll last week had the Presidential race dead even. McCain and Obama each had 48% of the vote.

Now, let’s go to the small print of this poll, its methodology. Rasmussen had 3,000 respondents. Each of these respondents were "likely voters", meaning they intended to vote in the upcoming election.

Rassmussen called these respondents between 5-9PM on week nights, from 11AM-6PM on Saturday and 1PM -9PM on Sunday.

Now as a point of consideration for all polls, during the week most of my die-hard Republican friends get home between 6-8PM. They work in the private sector. So if we take their middle time in coming home on the weeknights (7PM), they would not be available for 50% of the time when polls normally call during the week.

On the other hand, core supporters of the Democratic Party base, such as teachers, retirees and union workers have a higher probability to be home at these times.

In addition, I believe a person is more likely to partake in a 15 minute automated phone call when you get home at 4:30PM in the afternoon, then if you got home at 7PM at night. So there is some inherent bias in these times to begin with.

Now lets look at our second poll to compare. CNN this week has Obama up by 4 points. Four points ahead of the Rassmussen poll which was taken the same time. What gives?

Here's what gives. When diving into this poll's methodology it states that 1,020 respondents were polled, but 30% of them were not "likely voters" or not even registered to vote.

Compiling "registered voters" instead of "likely voters" always benefits the Democratic Party numbers returned and shades the results towards the Democratic candidate. Democratic voters are more likely to be registered and not show up to the polls than Republicans. That is why the most sophisticated polls use "likely voters."

So in conclusion, the Rassmussen poll had close to 3 times the sampling and 30% more "likely voters" incorporated in it. It is my opinion that the Rassmussen poll would be the one out of these two polls that I would look to.

In keeping how methodologies can screw up results, today the Washington Post came out with a poll that had Obama 9 points ahead. Bill McInturff, a pollster for the McCain campaign, points out that Democrats had a 16 point higher representation in the poll than Republicans, meaning it was 51/36% Democrats to Republicans (trust me on the numbers) when including which way the undecided voters in the poll leaned as well. This means they polled close to 1.5 Democrats for every Republican. Are they kidding? So goes the Washington Post.

In addition, 28% of the Post's respondents, more than 1 out of 4, of those polled were not "likely voters." Garbage In.... Garbage Out. No wonder the ridiculous result.

Now, I would not be surprised to see a 2-3 point swing to Obama with last week's financial crisis. But 9 points?

Com'on guys, polls are your livelihood and your profession. At least try to make it a somewhat fair.

As Mark Twain popularized the saying in America, "There are three kinds of lies: lies, damned lies, and statistics."