4 polls added today in IA, OH, MA & WA, with little impact on the map.
Today, I wanted to talk a little bit about margin of error. This morning, the Des Moines Register released a poll of 650 likely Iowa voters showing Obama in the lead 49 to 45 over Romney, with a margin of error of 3.8%. Most people would think that if that’s true then Obama’s lead is outside the margin of error. This is the great deception of polls.
What margin of error actually means is that with 95% confidence, it can be presumed that Barack Obama would receive somewhere between 45.2% and 52.8% of the vote in Iowa (+/- 3.8% on either side of the result). It can also be assumed, that Mitt Romney would receive between 41.2% and 48.8% of the vote. If this is true, then there are a considerable number of circumstances where Mitt Romney could be ahead. Certainly more than the “outside the margin of error” folks believe.
If we know that the 95% confidence level (+/- 3.8%) represents 2 standard deviations from the mean, then we can presume that 1 standard deviation is 1.9%. In a normal distribution, roughly 68% of outcomes are within 1 standard distribution, with 16% of the distribution more +1 SD and 16% less than -1 SD from the mean. If this is true, then we can presume that Mitt Romney has approximately a 15% chance of winning Iowa, if the election were held today, based on the results of this poll.
No comments:
Post a Comment