The WAL

As entertaining a tool as they are, polls are actually pretty useless.  I say this not to slander or annoy the people who make their living from them, but because it’s true.  They’re nice to identify trends and they give the pundits something to talk about every single campaign day, but they’re always wrong.  Aside from the ridiculous error bars (Exactly what does +/- 2.9% 19 times out of 20 mean, anyway?  That twentieth time is completely meaningless?  I know the math must bear things out, but come one guys, it’s like you’re admitting that you have no idea what you’re talking about.), they’re not accurate enough.

And what do I mean by that?  Well, you need to read the fine print.  The numbers usually add up to something in the vicinity of 100% which means that everyone they talked to actually thinks they know who they’re going to vote for.  So any given poll needs to be read as “XX% of decided voters intend to vote for party ABC”.  The undecided voters, by some accounts, can be as much as 40% of the total electorate (or the fraction of it that actually goes out to vote).  That’s an awfully big swing vote, don’t you think?

One more thing that bothers me about the math: can you really get an accurate picture of the way things are (even with a 3% error 19 times out of 20), by taking a sample of 1000 or so out of a population of millions of voters?  Seems to be an awfully small fraction to me, if there are, say,  10 million voters (this is very low - almost 15 million people voted in the 2006 election) and you ask a random 1000 of them what they think, you’re only getting one hundredth of a percent of the possible opinions at that moment.  The math may work, but people are more complicated than math.

Lastly, and this almost ignores my previous arguments, none of the polls are detailed enough.  A national picture is nice, but it’s not the whole story.  It’s lovely to be able to estimate seat totals if we all voted today, but it’s not the whole story.  Even going to the provincial level won’t do.  Something that really matters to me is how my riding is shaping up.  Is it a close enough race that I should vote strategically to not get the idiot I don’t want?  Or should I truly vote my conscience because there’s already a clear winner?  Yes, perhaps I should truly vote my conscience regardless of the front runner, but if it really is a tight race, then a little strategic voting could make a huge difference in my riding and the shape of the government across the country.

So my advice of the poling companies is this:  more detail, more clarity, more people surveyed.  I’ll answer a political survey every day if someone calls.  Not a marketing one, though - they can bugger off.