The Truth About the Polls: Statistical Error?

by Michelle Dong

When the results of the election came in, I was utterly confused. Some, like me, were stalking the polls since several weeksprior. The question was, how on Earth did Trump win the election, when all the polls were saying otherwise?

Of course, not realizing that 99% does not mean 100% was our mistake. Unlikely is still a greater chance than impossible, and likely is not the same as certain. Polls are predictions of the future, like a crystal ball, but are fueled by hundreds of statistics. But they are never completely accurate – after all, the Cavs won against the Warriors despite the polls showing otherwise.

Let’s look at the current NBA predictions on www.fivethirtyeight.com/. The likelihood that the Warriors win the title is currently 46%, while the Cavs are at a mere 15%. You might be thinking, that doesn’t add up to 100%. Why? Because it’s never certain, never 100% likely, until it actually happens.

Here’s the thing with polls. It’s the fact that they don’t add up to 100 that protects them. They can just rely on the other x percent that said the team wouldn’t win. In other words, if the Cavs win, the polls wouldn’t be “wrong” per se. It would simply be that, based on their polls, 15 out of 100 times, the Cavs would win. Does that mean that the polls made an error? Not really. It just so happened to be one of those 15 times.

Polls also take in the probability that the Warriors or Cavs don’t make the Finals, which is why the Spurs have a 11% chance of winning the title. Even though, technically, there’s only two teams that end up going against each other.

Statistics. Hate ‘em or love ‘em, may the odds are forever in your favor.

So, to answer the question:

Short answer – no.

Extended response – see article.