[javajosh]

It's interesting to view recent events in the US as a crisis of measurement. The news media measured the voting public with polling samples, we then measured the voting public with a vast government apparatus. Clearly, the measurements did not agree. Logically there are three possibilities: the media was right, the polls were right, or neither were right.

It's actually quite a good thing that we start to speak openly about threat-models against all voter sentiment measuring tools, especially the official ones.

[vehementi]

The last polls before the election put Trump at a 30% win chance. Trump winning is a completely normal outcome in that situation and doesn't signal polling problems.

[grzm]

There are two different types of polls we're talking about here. One are polls conducted during the campaign to capture potential voter sentiment and perhaps predict the outcome. The second are exit polls attempting to measure the election outcome based on what voters say they cast. (Nearly?) all complaints about polls I've seen have been about the former.

There are different sets of issues with both categories of polls. If anything, I believe the issue is not a measurement issue as much as it is educating the public about what the different types of polls are, their limitations, and their usefulness.

[theoh]

The idea that an electoral poll incorporates a random error (and maybe a systematic one too) is alien to the discussion about recounts etc.

[dmfdmf]

Yes, this was one of my questions. What is the accuracy and error margin of the voting system itself (and each state seems to have different methods so its not even uniform). Of roughly 126 million votes cast, 1% accuracy would be 1.26 million votes which would be of the same order of magnitude of Clinton's "win" of the popular vote. I have no idea if 1% accuracy is close or wildly too high or too low but we can't have a meaningful discussion without it.

Correction: Actual total was 126M votes for the presidential election not 160M votes as I stated. But the question remains.

[theoh]

Don't hold your breath.

Consider the case, in the technology of proportional representation, of the Hagenbach-Bischoff quota: https://en.wikipedia.org/wiki/Hagenbach-Bischoff_quota

It's (it seems to me) obviously arithmetically unfit for purpose compared to the Droop quota. In some cases more candidates can meet the quota than seats are available. Yet it remains in use/part of the credible discussion.