One of our biggest fans sent us this article, which cited an "expert" claiming that "Ohio's Vote Against Pot Legalization Was 'Statistically Impossible.'"
Statistically impossible? Really? Let's take a closer look.
- First of all, consider the source. The article came from Alternet - not the NY Times - so we'd want to verify the facts. (Even if it came from the Times, of course, we'd still want to see the proof.) For example, a Cleveland.com article said "only 43 percent of survey respondents said they were likely or very likely to vote for Issue 3" which, if you accounted for the margin of error in the respective polls, is not far from the 36% who actually did vote for it. So it looks like Alternet may have been cherry picking their sources.
- Secondly, the article uses polling data, which is not always 100% reliable (remember Dewey defeats Truman). With polls, you have margin of error, which expresses the level of confidence in estimates statistically. And you have to trust that the pollsters are truly taking representative samples. For example, a lot of younger folks don't have landlines, so if pollsters only survey people with landlines the results could be skewed - especially for an issue (such as marijuana) that could divide young vs. old.
- There is also prediction error - the idea that opinions can change over time and this may not be picked up in your prior polls. Sampling the entire population won't eliminate the fact that people change their minds.
These are just a few of the issues we found with just 10 minutes of digging. We're not saying alternet is wrong - but there are certainly some issues that need to be explored before you can claim that something is "statistically impossible."
(Thanks, Richard, for sending us this link.)