Monday, September 8, 2008

why i don't trust polling data

I'm not one to follow "polls" when it comes to elections, however difficult that might be for an obsessive news-reader.  I simply don't think they're accurate - and thank goodness for that.

My problem with constant polling data is that it can skew people's opinions about candidates when the only information it provides is "who is voting for whom."  I simply don't think that is good basis upon which to make a decision of who to vote for.

Even worse is shoddy record of accuracy.  Again and again during the US primaries, polls were at times radically at odds with actual results.  They seemed to get better as the election went on, but there were still noticeable discrepancies.

There are some possible reasons for this unreliability.  The one most discussed is the number of "cell-phone only" voters who are generally not polled.  This is because (so I've heard) some or many states have laws against unsolicited calls to mobile numbers.  As well, generally if you're receiving a call on a cell phone, you are on the move and probably don't want to spend a few minutes responding to a(nother) poll.

The thing is, this is a pretty specific demographic.  "Cell-phone only" (as opposed to "landline only," for example) voters are generally 35 or younger.  Not to mention they have a certain income brack and education level.

Now, polling companies have very complex systems for eliciting and analyzing data.  They likely consider factors such as the likelihood of certain demographics lying about their choice or not even voting come election day.  I would like to know how often they update these metrics.  For example, a few years ago "cell-phone only" voters probably likely to have a high income and good education, whereas now you're likely to see more low-income earners having only a cell phone because it's cheaper than a landline.

And the "cell-phone only" factor is only one of the issues with polling data.  What about people who screen their calls using caller ID?  What about people who are on the "do-not-call" lists? Perhaps the pollsters take this into account, but that would be a pretty complex algorithm just to elicit two percentages (red vs. blue).

So while I yearn to have up-to-the-minute status updates of the candidates' standings, I really don't think I can take this information seriously.  And that's great, because right now McCain has taken a slight lead over my man, Obama...

No comments: