Tuesday, October 5, 2010

Why I Don't Believe Any Polling Data

Math was my major at Temple University (1984).

We needed to take Probability and Statistics as a required course. Polling deals with trying to use statistics to measure or predict an outcome. The data you use to make your conclusions is the most critical part of the puzzle.


If you have garbage for data you get garbage results -- GIGO -- Garbage In Garbage Out.

I do not believe any poll can be valid at this point for following reason:
You cannot get a random sample of likely voters.

Years ago in the pre-cellphone era -- the pollster had the white pages as a universe for his sample. Know the exchanges and you could make the calls for an entire district. We all knew Havertown, PA had exchanges in the 610 area code (or 215 if you are older) or 446, 449, 789, or 853. If we were doing calls into there we knew right away who we were dealing with.

Exchanges and for that matter, area codes, are no longer geographically dependent with the ability to transport your number with you.

Do you own a cell phone? Is it listed? Does your college aged, registered to vote kid have a cell phone? Is it listed? How are the pollsters going to get to you if you aren't even on their lists?
Now, ask yourself, who still owns land lines? Older folks, right? Even some of those folks are losing them to become wireless. They are now unreachable.

Any polling to land lines are now skewed by definition. You are not going to get the younger voters. Unless you somehow get their phone numbers from working in another campaign, you can't get to them to ask them anything.

Even voter history files do not always contain correct phone numbers.

There is still a large and growing ocean of voters who are off-the-grid to the pollster. They may be registered to vote, but they are not active enough to give their numbers to anyone. No one can reach them to see if they are going to vote or not, much less where they are leaning.

Until bad sampling issues are rectified, the only poll that means anything will be the one conducted by the government in November. Until then, I think everyone is flying blind.

Honesty of Questions:

Many polls broadcast have loaded their questions. We all hear these type of polls:
"Do you think we should let the cannibal Democratic candidate anywhere near a morgue, or do you favor having armed Republicans guard our honored dead?"
This may sometimes be called a generic congressional poll.

Seriously, I know the questions really aren't that bad, but many times the subtlies of the questions key the respondent to the "correct" answer the pollster is looking for.

Are the questions neutral? How can you tell? You need to look at them yourself.

What to do?

One really needs to look at the polling sample and methodology combined with the actual questions prior to giving or reducing the credence of a poll. That is really the job of the media. (Yeah, I know the great job they do...) As a citizen, we need to question all the information coming from the boob tubes and internet.

For now, I am not putting too much faith in anything that has the word "poll" attached to it.

No comments: