USpol, presidential election, the polling 

People are wrong in saying the polls were terrible this year. They weren't great, but they weren't terrible.

Biggest takeaway: most of the polling errors were favorable to Trump, which means that the issues of 2016 were not fixed: pollsters are still failing to properly sample and/or adjust for conservative voters. That's something they will have to examine much more.

However, there's also the confounding factor of voter suppression by the Trump administration and GOP state governments, including slow-walking or disallowing voting by mail, sabotaging the USPS, and shit-talking voting by mail. Pollsters and forecasters, including FiveThirtyEight, made no attempt to adjust for that, because they had no idea how.

That said, election day went very smoothly and turnout beat records, so I doubt it was much of a factor in the end.

The attached image consists of the main bottom lines I took away from my analysis. My data is linked in my Dropbox.

All prediction and polling data (using their adjusted average) was taken from FiveThirtyEight.

USpol, presidential election, the polling 

@Garrison When you say "polling error favourable to X", does that mean the polls forecasted an erroneous advantage for that candidate or that the candidate got more actual votes than the polls predicted?


USpol, presidential election, the polling 

@JenLA it means the candidate won a greater percentage of the vote than the polls suggested

· · Web · 1 · 0 · 0

USpol, presidential election, the polling 

@Garrison Thanks, this makes the 2016 comparison make sense. Interesting stats! It seems like similar trends as in Europe, where the "nationalistic" parties usually poll lower than the election outcome.

Sign in to participate in the conversation

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!