[flagged]
For background on the suit and wiping of the data: https://www.apnews.com/877ee1015f1c43f1965f63538b035d3f.
That video is from 2014 well before the 2016 US election was a thing : https://www.youtube.com/watch?v=w3_0x6oaDmI
edit : interesting how in the recent comments under this video there's cryptoblockchain enthusiasts who have completely missed the point about lack of trust.
I would have thought that the way there election went down with multiple narrow state wins just above the recount level would have had at least some people wanting to take a second look.
There are two more common voter issues, that each amount to much higher errors. No voting system can accurately register the vote of an individual; the 2000 election in Florida was an example where the margin of victory was on the order of this inaccuracy, one of the reasons its result was so contentious. Furthermore, there's good old voter intimidation and similar techniques: tell people the wrong polling place, deny them the ability to cast their votes, etc. I don't know the prevalence of the latter case, but it is far more than 1 in a million, based on evidence.
What do you mean by this? Do voting machines have an error rate? I would assume the county would be notified and these ballots counted by hand during a recount (which is what happened in 2000).
I think 2000 was contentious because the Florida AG stopped the recount as Gore's numbers were steadily improving and threatening to overtake Bush's.
Another possibility is that labels and voting regions might not line up properly. The infamous butterfly ballots of 2000 had the property that, if you viewed from the appropriate oblique angle, the arrow next to Al Gore's name very neatly lines up to the hole next to Pat Buchanan. Unsurprisingly, in Palm Beach, there was an unexpectedly high vote count for Buchanan in a region which most observers would think highly unlikely to vote for him.
What happened during the 2000 recount was that Gore asked for a recount of specific counties that he felt would improve his tally; the Bush campaign countered that all the counties should be recounted, since recounts in other places might improve his tally. The Bush campaign also argued that the varying standards of what constituted a vote (did you need to have at least two corners of the chad hanging, or did you merely need a dimple, to constitute a vote?) also was too lenient a standard for the recount.
The general consensus of most observers, after the election furor died down, is that Al Gore would have won the state under a 100% accurate vote counting mechanism. A full recount likely would have been to Gore's favor, but the recounts that Gore asked for do not appear to have been sufficient to get that result.
When the results of the election change depending on the exact measure you use to determine the intended vote, it is fair to say that, in a statistical sense, the election is a tie. I would argue that election laws should treat such cases as a tie, even if the numbers aren't actually identical.
I didn't know that Gore asked for recounts in specific counties while observers thought a state wide recount would have helped him as well. What a blunder...
https://www.theroot.com/evidence-shows-hackers-changed-votes...
There doesn't seem that much enthusiasm for investigating. The first hard evidence that "Russian government coordinated a spear-phishing attack on computers at an American voting machine company and compromised at least one email account" was leaked in June 2017 by Reality Leigh Winner who was prosecuted for leaking that and is now in prison in Georgia.
I think we can all imagine that computers can be programmed to perform this logic fairly trivially.
We can all imagine that computers can be hacked the way you describe, but such computers don't exist everywhere a hacker would need them. Take Michigan as an example, since it was key to Trump's victory and only a handful of counties swayed the results. The entire state uses paper ballots. How can this system be hacked without a massive social engineering conspiracy involving hundreds of people?
https://en.wikipedia.org/wiki/Facebook%E2%80%93Cambridge_Ana...
Everything else you're saying is spot on, though.
The average of the 13 final national polls had Clinton ahead by ~3 points, Clinton won the national vote by ~2. Some individual state polls were more wrong, but they were conducted less frequently and so many of them didn't fully factor in the latest news cycle developments.
Even if we exclude mendacity and downgrade that to mere irresponsibility, terms like "98% chance of Hillary winning" were being thrown around left and right on election day. That speaks to something very, very wrong in how that data was being collected.
The problem was mostly with how it was analyzed, not how it was collected. Some publications were making bad assumptions that state by state polling errors would end up being independent. 538, who didn't make that assumption, gave trump close to 1/3 chance of winning.
Trump won by less than 80k votes in 3 states. Even the state polls weren't that far off--unlikely events happen.
The polling was fairly accurate; given the nature of the electoral college and the fact that state-by-state deviations from polling results tend to be correlated rather than independent, and how close many of the state contests were, the winner was different than the expected winner (and this was very surprising to people who followed predictors who based their odds on the assumption that state variation was independent, less so those who followed those like 538 that pointed out that that was a mistake), but the polling was not particularly inaccurate.