Every day we hear new numbers, shifts that fall within the margin of error. How much sense does it make to report that the NDP has increased their support to 16% from 15% if the margin of error is 2%?
This is the best thing I've read so far on polling for the federal election: Rob Annandale at the Tyee posted a story about two different approaches to election predictions: a results stock market and a gossip poll. Between them and the pollsters, can you guess who was more accurate? You're right, it wasn't the pollsters, it was the million monkeys typing on a million typewriters. Or something.
Two anecdotal proofs*:
1) Every year the Globe and Mail's Business section has a stock-picking competition among five or so analysts. They each pick one stock to perform well over the year. They also give one pick to a random child of an employee (changed in recent years to include high-school students). The year I read the results, the five-year-old daughter of one of the editors who picked the results with a doll placed third out of six.
2) During the NHL play-offs the team at TSN has a monkey pick play-off results each round. Maggie the macaque's picks have been eerily prescient on occasion. She has been right more often than not.
*that may or not be factually correct. Change anecdotal to apocryphal and I'm right as rain. Or something.