Poll Averaging is Science but most Sanders Supporters Don't Use It

For the past five months, my Facebook news has been full of pro-Sanders friends posting stories about how well Sanders is doing in the polls. They systematically cherry-pick the most outlying polls, while ignoring the outliers that go in favor of Clinton.

For instance, they'll pick the recent NH poll that had Sanders 60% to 33% for Clinton, but ignore the two polls that have the race at 46% to 43% and 49% to 43%.

With some exceptions (the UK Conservative majority in the last election), we're now in an era where most election outcomes can be predicted very accurately by poll-averaging. This is especially true for larger elections with more polling and for elections where one candidate (Clinton) has a massive 17% national lead over the other.

Everyone should use poll averaging. For instance, I recommend the Huffington Pollster

Now Sanders may have a small chance of winning as he has some momentum in the national polls (and at a linear growth rate he might tie Clinton's national support in March to April), but he still has a long way to go.

If you want to get fancy, you can add additional criteria to your model. For instance, 538 is doing this and has two different models (poll averaging and a poll plus model). This poll plus model, which heavily discounts Trump and Cruz is probably just as likely to fail as it is to succeed. It would require a lot of back testing to get it to work. It might be working similar to a prediction or betting market (like Predict It, in which case there is some hope for a better result.

The poll cherry picking is encouraged by social media, the mainstream media, and by the campaigns.

But if you believe in using science for things like climate change, you apply it consistently with polling as well - otherwise you are undermining your credibility.