> the algorithm was said to play a role in virtually every strategic decision Clinton aides made, including where and when to deploy the candidate and her battalion of surrogates and where to air television ads — as well as when it was safe to stay dark.
Meanwhile, Trump was doing 3 rallies a day, across multiple states, and up to 5 a day (including until after midnight) in the days leading up to the election, drawing crowds in excess of ten thousand people.
Where was Clinton during all of this? Almost nowhere to be seen. It seems to me that Ada made some pretty terrible choices.
I don't know about that – now that she's lost every strategic decision can be made to look wrong. But the real problem was that people just weren't drawn to her as a candidate. The algorithm isn't to blame.
And maybe having her hold more rallies would have worked against her in that case (people not being drawn to her) because there would have been the an increasing emphasis made on the comparative sizes of crowds (which was already happening to some degree).
That puts them in between a rock and hard place strategically - hold more rallies and get lambasted for having smaller crowds, don't hold the rallies and get lambasted for not having the stamina and not being accessible to the people.
In these cases, one wonders if a human in charge of strategy might have made better choices. After all, it was only just recently that a computer was able to outmatch a human in Go (heavily based on strategy) and I'm not convinced that an election involves less strategy and that you can rely on an algorithm to beat a talented human opponent.
And say what you will about Trump, but his strategy this entire election campaign has paid off incredibly for him and he did what most people considered to be the impossible.
I don't know if "his strategy paid off" is correct - it seems to me that at one stage he wanted a way out and actually tried to fail, but the people just wanted him to go on. What's that quote about him shooting someone and that not mattering?
He could well have been elected despite his strategy, not because of it.
> it seems to me that at one stage he wanted a way out and actually tried to fail
Which stage was this? I mean sure there were plenty in the media quoting unnamed sources stating that Trump was going to quit any day now - for almost the entire duration of his campaign, in fact.
There were also plenty of times when he went against conventional political wisdom and came out on top. I don't think that was him trying to throw the campaign, I think it's just a fact that he's not a politician and so he does things differently.
Trump was always very clear that he was serious and that he planned to become president.
I think thats it entirely. I'd be very curious what their model had if anything for rural/urban voters. It could be that even if she knew there was very little support amongst rural voters, she as a candidate had no way at all to pivot her strategy towards these voters without alienating doners and her base.
Trump ran as the face of rural voters - every attack Clinton made became an attack on people who agreed with him. Perhaps not agreeing with every single outrage, but general statements such as "he's just a racist" were something that the rural voter had heard over and over again as a way to dismiss them. Trump turned the attacks into a provocation of his base.
This isn't even a new strategy, the exact same thing happened with GWB. Electoral college victory and everything. The DNC are incompetent hacks to not have seen this coming.
Looking at the rust belt states that Trump managed to peel off, I think she had two things working against her:
1. Her policies were less vivid and memorable. Her tax policy versus Trump's belligerence at trade partners Mexico and China where the factory work moved.
2. Her credibility regarding her ability to actually bring about change. I think Bill Clinton tried to address this in his speeches, but she is part of the establishment and therefore expected to continue the status quo.
This is apart from her health, which I think affected her ability to fight the media war, and being a woman, which makes her an icon for social issues whether or not she wants to be -- those maybe a factor but they are outside her control.
Definitely, especially with the media being ordered to do the same thing. All you could hear on corporate media is "black people" and "white people" and "latinos". You'd hardly ever hear words like "working class people" or "blue collar".
It was all about race and gender. Well guess what? Most people didn't vote based on gender and race issues. They voted on economy and corruption (which is related to the economy one) issues.
Meanwhile, Clinton's entire campaign was about how much of a bigot Trump was, while half of the country earns less than $30,000 a year.
Who campaigned for this half of America? Well, it wasn't Clinton. She campaigned on "4 more years of Obama" and "not rocking the boat or changing much." Those would be pipe dreams, after all.
Only Trump and Sanders did, and Sanders they cheated out of the election in a multitude of ways, including by spreading the Democratic establishment propaganda that Clinton was the "more electable" one because she was "battle tested" (she didn't do so well in the last battle, though). But she couldn't cheat Trump out of the election, so he won.
The worst thing about this is that the Democratic establishment don't even want to accept the blame for the complete and utter loss of all branches of the government. They want to coronate Tim Kaine at the next election - the guy that nobody knows anything about, except that for some reason Clinton picked him, and also part of the same team that has just lost. Great strategy. I can already see all the people that didn't go vote for Clinton not come to vote for Kaine in even greater numbers!
2. The system selected Pennsylvania as an important target.
Apparently it wasn't the choice of PA as an "important" target, but overfit of its importance to that of other key states (e.g. in the rust belt) which may have been at fault. Like the article says:
But it appears that the importance of other states Clinton would lose — including Michigan and Wisconsin — never became fully apparent or that it was too late once it did.
Looking at the results, I can agree that focusing on Michigan would have been more optimal, but to win I think she needed MI+WI+PA, so if she overfocused on PA and still didn't take it, broadening her effort may have lessened her loss but I don't think it would have been a path to victory.
I'm pretty sure the data collection was not well handled.
Why would we think that the hubris, as exposed in the Wikileaks emails, would not also infect this "secret" project.
I also wonder if everyone involved was able to speak truth to power--Or if the ones with dissenting opinions or countering data were simply ignored? Seems plausible.
Big MCMC model, likely written in Stan. Fancy as can be to a statistician with a penchant for strong priors... And completely fucking useless when your priors on likely voters are as wrong as wrong can be.
I think it's pretty clear that when a fair portion of your data (polls) are misleading or incorrect for whatever reason - your algorithm will have a very hard time producing meaningful output.
I think it did work. The Clinton camp uses internal polls which are much more accurate than public ones. These don't have the 1-2 week lag that the public polls do. The Trump campaign has its own internal polls too. From what I've read, the Trump campaign's internal polls found Clinton support in the Midwest collapsed 1.5 weeks before the election. From how they reacted it sure looks like the Clinton camp knew this too. The Clinton camp suddenly went all out: surprise rallies in Michigan and Pennsylvania; surprise concerts with Jay Z, Madonna, Lady Gaga; even drafting Obama to do daily rallies in November. Obama held twice as many rallies in November than he did in all of October.
The way I view it the Clinton camp knew they were in serious trouble and reacted accordingly.
She was also campaigning in Arizona the week before the election, a state that would be really hard to win, and not campaigning in Wisconsin, where she had a chance. So not sure they did have the best picture of what was going on, or at least didn't respond to it.
If the collapse on the most correct polls happened at that time, then could it not be seen as an election swinging event? Not saying Hillary Clinton and/or the DNC are not responsible for being weak enough to be able to be routed, but its an example of... well, the FBI significantly turning an election.
By that point Wikileaks had also been releasing Podesta emails for about a month. It's an example of Clinton's terrible decisions re: the email server turning an election.
Maybe it did work? The loss may have been even greater without the data driven decision making.
There's been a ton of finger pointing today as to where the blame falls for Clinton losing (Facebook...). The reality isn't hard to understand, Bernie was treated unfairly by team Clinton and Trump's supporters were discounted, vilified and talked down to. It's pretty easy to see where the real fault lies and it's not with any form of technology.
I'm reminded of Mitt Romney's fabled ORCA system[1], which was also an epic failure. Indeed, it's something which has since caused me to re-evaluate my opinion of Romney: if he was as great an executive as I thought, how could he have helmed a disaster like ORCA?
You're all going to hear a lot more about the Shy Tory Effect[1] in the coming months and years. It was first identified in Britain. It's talked about in Australian politics too.
In particularly heated elections, right-leaning voters begin to lie to pollsters about their voting intentions. They are embarrassed, or afraid, or skeptical about the pollster's intentions and honesty. So they lie. This creates a misleading impression of the actual position of the candidates.
Typically, exit polls show a much more realistic picture of the vote. It's just that they're too late to operate on.
The next generation of algorithms will need to introduce allowances for the shy tories in the US, relative to the degree to which one side feels unwilling to be honest with pollsters.
My understanding is exit polls are actually pretty unreliable; they are conducted primarily to correlate demographic data with the official totals for future campaigning purposes, but need to be adjusted based on the actual outcome as the raw data is not representative. Perhaps this particular phenomenon is less present, though?
I think I see the disconnect. I'm Australian, we have compulsory voting. I hadn't connected all the dots on the effects of voluntary voting on exit polls. Only got as far as thinking about the ballot.
> In particularly heated elections, right-leaning voters begin to lie to pollsters about their voting intentions. They are embarrassed, or afraid, or skeptical about the pollster's intentions and honesty.
You are in the right direction, but I think it's a bit not like that and not about lying, being afraid and being skeptical about pollster's intention.
If people feel shame for their choice, they would try to avoid expressing it anywhere, participating in polls, etc., especially in front of their peers. In regions where most have opposing views and shame for choosing the other candidate, they will feel pressure to keep their choice to themselves and they won't in the regions where most support the same candidate. Being afraid of something is a completely different thing and generally prevents voters to even vote, not just avoid polls.
Britain has voluntary voting too. The phenomenon has been observed in several elections there and in Australia, where voting is compulsory. So discouragement of voter turnout is neither here nor there in the effect.
In the end, it doesn't necessarily matter why it happens.
The other factor is that the polls were likely rigged in her favor since she was spending so much more on paid advertising. Note how they all magically converged on the last day of the race.
> And for the journalists working in the next room to not want to win a Pulitzer for exposing it.
Has there ever been a case of a journalist winning a pulitzer for uncovering corruption within their own newsroom? I thought there was a de facto ban on that. (E.g. I don't think the NYT won a pulitzer for reporting that they had previously censored all news about Bush's spying infrastructure to help him get re-elected.)
If the discrepancy is primarily due to the shy tory effect and not the polls being rigged, this should be easy to see from the data because robopolls should be vastly more accurate than live polls. Is this the pattern that actually exists?
edit: "The study found only a slight impact by moving poll respondents from the internet to a phone call with a live interviewer — with larger effects among college-educated white voters."[1][2]
So while I think it's likely there was a shy tory effect, the evidence suggests it was small enough that there must have been other factors at play.
> this should be easy to see from the data because robopolls should be vastly more accurate than live polls
The same pattern is observed in other countries regardless of polling method or polling company.
So unless Clinton has been rigging polls world-wide for decades in anticipation of her 2016 run, it seems more plausible that people don't trust pollsters.
I'm not sure if I'm supposed to be excited about the name: "wow, their programmers know about Ada Lovelace," or about the fact that campaigns are high tech, which isn't new but I guess is still a surprise to some voters.
It seems like the main problem is that no one took into account the true effect of the 2011 gerrymandering and assumed the support for Obama in 2012 could still be relied upon as a predictor.
can you explain exactly how a presidential election is (or could be) gerrymandered? Aside from two (Iowa and Maine) all the states award all of their electoral votes to the winner of that state's popular vote. So unless the shapes of the states change, there's no way to gerrymander.
Meanwhile, Trump was doing 3 rallies a day, across multiple states, and up to 5 a day (including until after midnight) in the days leading up to the election, drawing crowds in excess of ten thousand people.
Where was Clinton during all of this? Almost nowhere to be seen. It seems to me that Ada made some pretty terrible choices.