October 20, 2006
Statistical Ties
In politics, polls are everything. One thing you always hear is when one politician trails another by 5 points, with a margin of error in the poll of +/- 3%, all the pundits who favor that politician call it a “statistical tie”. After all, if you swing one guy up by 2.5 points, and the other guy down by those 2.5 points, you’ve got a tie. Now, it doesn’t matter if 12 different polls all find the margin a consistent 4-6 points, as long as each poll has the same margin of error, the pundits call it a statistical tie.
Now, if I poll 1000 people, asking them today, “which party do you think will win the House”, and I get a 5-point swing, with a margin of error of +/- 3 points, you can take that poll as meaningless. However, if I, and three other polling firms get the same result, the consistency of result means that the chance that all 4 polls are wrong by 2.5 points in the same direction is fairly low. So four polls that show someone is down 5 points pretty much means that the guys is down about 5 points. While each poll has a +/- 3 point margin of error, the combined result is probably, at best, a +/- 1 point margin of error. The more polls showing the same result, the more likely it is that this trend is real, and not an erroneous poll.
Of course, that’s no reason to believe polls at all. I much prefer to believe prediction markets like tradesports.com, as the old adage, “Why don’t you put your money where your mouth is” is answered there by people actually doing it. But I’m also a big fan of pointing out idiocy of the media, and the pundits take the cake. One poll may be a statistical anomaly. Multiple polls, all agreeing with each other, are something to take notice of.
No Comments
No comments yet.
RSS feed for comments on this post.
Sorry, the comment form is closed at this time.
