Friday, May 17, 2013

Four Factors explain Declining Reliability of Election Polls

(1)Growing voter volatility means, other things equal, there is a greater margin of error; (2)lower turnout means you have to determine whether voters surveyed intend to vote; and (3) changing technology means that land lines are either harder to access or less representative as a sample when they are reached.

It is (4) the fourth factor, explained in this article in the Vancouver Sun by Jooan Bryden, that is perhaps least well known or understood; that is the discrepancy between more sophisticated and fine-tuned survey techniques and more standardized methodologies that public opinion conducted for media organizations tend to believe in.  What fooled me in last Tuesday's election was that I had been reassured because I had heard that pollsters were doing online surveys that were confirming their other data; Tom Barrett's article  in The Tyee explains the difficulty with online polling , in terms of self-selection and representativeness. 

What is needed is some kind of methodology for integrating online, cell  and land line so that they compensate for each others' weaknesses. Easier said than done.  But more to the point, media and NDP will need to match the more well-heeled Liberal and Conservative campaigns  in conducting intensive, non standardized interviews and other methods. The simple fact is that, like weather forecasting,  standard polling is becoming less accurate rather than more so, due to very large structural factors that are difficult to fully comprehend and manage. But that doesn't mean that we should stop polling any more than it means that we should stop watching weather forecasts.

No comments: