Stupidity, plus or minus 5 percent
Sunday, September 14th, 2008 By Mike MaurerA poor fellow has the thankless task of teaching me statistics, and he pointed out this silly story from CNN, purporting to combine various polls to come up with a “poll of polls.”
Here’s the best part: “The poll of polls does not have a sampling error.”
Yup, plus or minus zero. That’s some accurate poll.
The best part about this is that this statement is absolutely correct. There is no error, because 100 percent of the population was sampled. Yep, all five of five polls included in the population.
Of course, not one reader in 100 or even 1000 understands what error means. Maybe 1 in 10,000 does. And that one knows that it has nothing to do with what most people assume it means, which is to say, how likely is it that the poll reflects reality? Margin of error is just a mathematical construct that isn’t all that important-which is to say, it’s plenty important, but it’s rarely violated, so to speak of it as if it’s conveying any real information that is relevant from poll to poll is misleading.
The issue isn’t margin of error. MOE *assumes* randomness, representativeness and validity and reliability-things that aren’t really known and can’t really be measured. And they’re much more important when trying to sort out if your poll is really telling you anything.
But hey, if you repeat a number, then things must be good, right? I liken it to significant digits: If it sounds impressive to say the budget is $32 million, then it must be much more precise and intelligent to say it’s $32,516,242.33. And the sad part is, many people really believe that to be the case.
Thanks to CNN for exposing the numbers fraud that polling stories invariably practice.
Tags: polls
