Lies, Damned Lies, and Daily Polls

If there's one thing I've noticed this election cycle — other than an increased personal desire to yell and throw things at the computer monitor — it's the love that certain pundits and media organizations have developed for daily poll results; Less than three weeks in and I've already seen pages of prognostication and blathering over everything from tiny national shifts to larger regional shifts.

Or perhaps I should say perceived shifts.  Because there's one thing I don't quite get about the obsession with these polls: the number of people surveyed per day tends to be rather low, resulting in large margins of error.  Of course, this means that any perceived shift has to be monumental in order to be confirmed as an actual shift in voter intentions as opposed to just potential error.  In other words: a 5% gain when your margin of error is +/-2%? That's a notable change.  A 2% gain when your margin of error is +/-5%? That could just be phoning the wrong couple of households on a given night.  (I saw a recent one where the margin of error for Atlantic Canada was 0.1% shy of a full +/-10%.  Somebody would have to eat a baby on live television for any change to be noticeable against that.)

On the plus side though, at least most of the numbers tend to include regional breakdowns.  Whether a 10% national support is concentrated in one region or consistent nationally makes a huge difference with regards to whether it translates into likely seats or not.