Sunday, July 4, 2010

Poll Silence - Potential Formula Change

It's been 10 days since the last federal voter intention poll was published. I'm beginning to suffer from withdrawal. How about you? :)

More seriously, if polling remains this slow, I may use an alternative formula for depreciation. Right now, a poll loses all value if its midpoint date is 27 days or more earlier than the midpoint date of the most recent survey (less than 27 if the firm has multiple recent polls). We know that EKOS will publish every two weeks. If, in addition, Harris-Decima, Nanos, Angus Reid and Ipsos Reid publish every month, and their polls are well spaced out from each other, I won't have to change the formula. But if these four firms publish results less than monthly, or if their polls come in lumps, I may have to lengthen the 27-day poll life span in order to prevent EKOS from dominating projections.

Whatever happens, I will revert to the current formula in the fall. When a campaign occurs, the 27-day period would most likely be reduced to just 10-13 days.

No comments: