Interesting Links for 10-11-2016

[andrewducker]

Why were the US polls wrong? tl;dr - people voted who normally don't.
(tags: polls usa politics ) VR Is Totally Changing How Architects Dream Up Buildings (tags: architecture VirtualReality ) Psychological safety is crucial for great teamwork-but it has a dark side (tags: psychology teams ) What does history tell us about what may happen next

Comments

[lil_shepherd]

I suggest that in the case of Brexit, it is indeed true a lot of people voted who normally do not. It was the highest percentage turnout for a national vote on anything for some years. Certainly higher than the General Election (which was also a surprise result, and also had the highest turnout for a General Election in recent years.)

On the other hand, the opposite it true in the US. The turnout was way down on both of the two Obama elections. The Republican vote appears to have held up well: the Democrat vote did not.

[a_pawson]

This. The republican candidate got roughly the same number of votes as they always do, but the democrat vote has been falling over the past 3 elections. For whatever reason Hilary didn't inspire enough people to vote for her.

[kalimac]

Do you have this chart for a longer period of time? The impression I've gotten from non-statistical reading is that 2008 was an outlier, and I'd like to see that confirmed, or not.

But this is generally correct. I've seen a a few people convinced that the Green vote cost the Democrats the election the way it did in 2000, but though it was reasonable beforehand to expect that might happen, in the end it did not. All the Green vote going Democratic would have turned Wisconsin, but no other state, and that's not enough. Adding much of the Libertarian vote to it as well would have turned enough additional states, but as Libertarians are much less likely to be dissident Democrats than Greens are, it reverts to a turnout problem. "If people who didn't vote at all had voted Democratic" is a far more reasonable alternative than "If Libertarians had voted Democratic."

[skington]

Or, possibly, voter suppression tactics by Republican governors were to blame for the reduced Democratic turnout.

[danieldwilliam]

I have some thoughts on Trumps first 10 days but they are too long winded for a comment.

http://danieldwilliam.livejournal.com/150492.html

[steer]

Were the polls wrong? I would say they were comfortably within the margin of error if properly analysed. Five thirty eight had a 70% chance of Clinton victory -- roughly 2/3. I don't know why anyone had been reading that and said "it's a sure thing". If I say there is a 2/3 chance of scoring 3 or more when rolling a die I am not wrong if the next roll is a 1 or 2.

All the effects in "where the polls were wrong" articles are well known to pollsters -- different people show up, people are shy, people state the wrong thing, it's raining that day, some people change their mind at the last minute and so on and so on.

In this case if you increase the Clinton vote by 2% the 538 model is pretty much correct (I think they said 48 out of 50 states correct). Being within 2% is pretty damn good IMHO. It could be that if the vote was the next week or next day it would have moved 2%.

So I would say, for the poll I followed most closely, the 538 poll, they got it pretty close and were very careful about the margin of error they provided.

[steer]

And, in fact, you link the 538 analysis later in your list so I should have read your articles all rather than commenting on the first. :-)