Friday, October 24, 2008

Save Democracy - Lie to a Pollster

So we already know we can’t trust politicians. But do we dare believe polls now or not? Obviously we can’t rely on them as they become less and less reliable. According to a great piece by Siri Agrell in the Globe and Mail today, the American pollsters are all over the map on tracking opinions on the road to the White House.

Pollsters are showing results from 13 different pollsters we see claims of an Obama lead ranging from 14 points to 1 point. Can they all be right? Can any of them be right? The polling methodologies are the culprits. The old-school techniques are missing major portions of the population because they don’t connect with cell phone users…lots of youth are not in the results.

They use automatic dialing systems and lots of folks will not talk to a machine. More built-in inaccuracies are “normalized.” Some only use Internet pull techniques and then apply variables like multipliers to try and adjust for skewed data.

Many national polls have very small sample sizes for regional reporting but that does not seem to be any deterrence from pollsters still making statements about regional findings. Other efforts to account for wide ranging differences is be averaging all polls, the so-called poll of polls. This is still no way to get better polling accuracy because garbage-in/garbage-out is not resolved by averaging all the garbage.

Canadian pollsters were all caught in the 2006 election predictions because they failed to recognize the impact of last minute shifts and final impulse decisions in undecided voters. This election they did rolling polls over the weekend before Election Day and came up with much closer results.

Mistrust of pundits, politicians and now pollsters by the public is a prudent default position for anyone who wants to know what is really going on in politics today.