But was it fairly small in statistical significance? A polling error of 3% is likely big enough to confirm a consistent and very significant bias.
It's like throwing a coin and getting 52000 heads and 48000 tails. "Only" a 4% error, but sure as hell evidence your coin is biased.
What's perplexing to me is that the 3% error occurred in a polling average of dozens, perhaps hundreds of polls. If it were just one poll, a 3% miss is quite within the realm of possibility. But in this case it was a range of polls that centered on Clinton +4% and itself had a roughly Gaussian distribution between Trump +2% and Clinton +10%. This makes me wonder if the polling samples were themselves not independent, or some other form of systematic bias amongst mainstream pollsters.
https://twitter.com/samwangphd/status/788544053059153924
http://edition.cnn.com/videos/tv/2016/11/05/poll-expert-i-wi...
http://election.princeton.edu/2016/10/10/some-secrets-are-no...
Specifically the last link exhibits the smugness this article is talking about. Here's a quote from it:
> This seems like a good time to reveal one of the Princeton Election Consortium’s own secrets. Thankfully, it does not involve an Access Hollywood video. Here it is: poll-based Presidential prediction is not very hard. I guess that is a pretty boring secret. Sorry.
Sam Wang was predicting a 95% chance of winning for Hillary. I guess we're in the 5%. I'd further like to point out that Hillary is right now winning the popular vote by 0.1%, and the NYTimes has adjusted their prediction down to 0.5% when all the votes are tallied.
Obama's win over Romney was slightly less than 4.0%, and he won 332-206 in the electoral college. Here's how Sam Wang describes it:
> President Obama’s re-election in 2012 carried even less suspense: he never lost the lead to Romney. The closest he came was right after the first debate, though even then he was slightly ahead.
The pressure to express a Trump sentiment is not high in a non-swing state; If you are in an entrenched blue state, people do not care about you(guaranteed democrat win anyway). If you are in an entrenched red state, then you have plenty of support around you. The voters in swing states are a much smaller percentage of the total, and preference falsification would have affected those states the most.
In my last paragraph, I tried to put forth a theory that leaves the preference interpretation valid. What do you think about it?
I don't buy this. Sam Wang had projected Clinton winning the popular vote by 4.0% (+/- .6%). She's currently projected to win the popular vote by about 1% (the Electoral College margin is an artifact of the electoral system and unrelated to voting preferences). That means that maybe 1 in 60 Americans voted differently from what the polls predicted. So, while the margin of error relative to the margin of victory [1] was pretty big, the margin of error relative to the electorate was fairly small. It doesn't seem that a whole lot of people were hiding their actual preferences prior to the election.
[1] "Victory" only in the narrow mathematical sense of winning the national popular vote, obviously.