I went through a mild earthquake a few years ago here in Maryland and it permanently changed my perspective about how insignificant we are relative to the larger forces abroad in the universe. I understood earthquakes intellectually and conceptually but riding one out, even a tiny one was an eye opener.
I kind of feel the same way after the electoral earthquake Tuesday night. I was never a huge Clinton fanboy. Clinton redux after two Bushes was getting a little too entitled and imperial for me and she has (for me) almost zero charisma, but she won the primaries, and so I got in line and voted for her Tuesday afternoon.
Now I don’t know what to think. All the aggregating polls and confident media assertions I relied on as hard data and the basis for a solid rational perspective were liquefied in the results. The authorities I took at face value were all wrong. Everything I thought could be relied on was in shambles. I could be angry for getting played but it’s pretty evident the experts were not really operating in bad faith, they thought they were right too… until they weren’t. It’s slightly reminiscent of the housing collapse of 2006-2007 where all the confident experts were “right” about the solidity of residential real estate until the bottom fell out.
So… where do you go from here as a consumer of news? If analyses backed by billions of dollars in technology, manpower and expertise shits the bed who can you trust?
What real difference does it make if the polls are right or wrong, and just who is being conned here? Deciding who to vote for should be based on an assessment of the candidates’ qualities and should not be swayed by perceived popularity. Yes a lot of people were shocked by the outcome of the election and are upset, but once they are over that shock are the next four years going to be any worse because of it?
I wouldn’t say data is dead. Companies, researchers, etc. are still using big data and analysis for value-added purposes. If data was useless, it would be discovered quickly, especially in profit-driven areas. However, the issue with data is, and has always been, that something structural and widespread may just not be being captured. In the case of the election, everyone knew that there was a chance that Hillary could lose one or more of the swing states. But the conventional wisdom was that she couldn’t lose all of them and thus would still win. However, it turns out, that she lost each swing state (especially the Rust Belt) due a closeted (almost underground) group of Trump supporters that weren’t being captured in the polls and/or an unmotivated group of Clinton supporters that couldn’t be bothered to make it to the voting booth. Perhaps, in the case of the closeted Trump supporters, they chose to hide due to being too embarrassed to publicly support him. Ironically, Trump’s best strategy was being such a rogue that his many of his voters didn’t want to reveal themselves. This prevented Hillary from allocating her resources effectively.
You have to take data and experts for what they’re actually saying.
Using political polls as an example, how often does a media source actually show the margin of error? I mostly watch CNN and they often omitted the margin of error entirely when showing graphics that said Hillary 48% and Trump 44%. They also tended to assume that it represented current information.
But that’s not what the poll said at all! The poll said Hillary was between 44% and 52% popularity, two days ago, among likely voters. The election happened after the poll, and the election measured actual voters, not just likely voters.
And, yet, in defense of the polls, Hillary losing was a statistical possibility in virtually every poll. It’s like if I tell that you rolling two dice is most likely going to yield 7. I have made a true statement, but someone who understands statistics knows that about 80% of the time you’ll get something other than 7. The problem is, the media reported “This next dice roll is going to be a 7.” No, that’s not what the science or the data said at all and you cannot take the information out of its statistical context and pretend like it is still useful information.
Regarding the election, the real lesson is that people need to get off their butts and vote. Which isn’t new information at all. We’ve been telling them that for 200 years.
About investments: much the same thing. When we say “Stocks average a 6% return” what we really mean “Stocks have a distribution curve, which goes from a 100% loss to an infinite return.” So I would argue that the real experts were not wrong about 2007. The real experts know that there’s a 1% chance in any given year that you lose 50% of your investment value. I think the surprise is that maybe it’s more like a 2% chance of losing 50%. And yet, it is also true that summing up all the losses and all the gains comes back to a 6% on average. You just have to understand that the 6% average is wrong 90% of the time. It’s the center of a distribution, not a precise prediction.
In general, as a consumer of information: I accept that nothing is error proof and I do my best to understand the range of results covered by the statistics. I remember that you don’t round 5% chances down to 0% each year, you round them up to 100% over a 20-year period.
I read the headline for the OP and was prepared to get all angry at the overreaction it seemed to imply. The post itself was thoughtful and nuanced, though! (Maybe a lesson there about what journalists go through, actually. Journalists don’t always write their own headlines, but they often get judged more on the headline of a piece than its content.)
I’d draw a parallel to the scientific method. Hypotheses are tested, theories are refined, and we get comfortable with a consensus view of how things work. Then something happens and we re-evaluate.
This also reminds me of a pattern I’ve seen in my opinion journalism diet the past two days. (Slowly getting back into it.) Even very hard-core lefty writers have been willing to admit their surprise at the results AND start to reconsider some very basic assumptions they hold about what the left/Democratic party should do going forward. Example here: The Aristocrats! – Lawyers, Guns & Money. I also hear this kind of soul-searching going on with IRL leftish friends/acquaintances.
To me, I think it’s great that people are approaching this surprise and shock with open minds, instead of retreating into “IT’S RIGGED” - style denial. Hopefully, we’ll learn something, and turn some of these lemons into some kind of lemonade.
What is the opposite of threadshitting? I have nothing of value to add except to say that I love this thread and the thoughtful, nuanced responses so far. Great job, everyone!
It’s not necessarily just the data; the whole field of political consulting ought to be taking a hard look at itself just now. Clinton spent vastly more than Trump on conventional infrastructure like the ground game, GOTV efforts, TV ads, and polling, and it didn’t help. Her turnout was down and poorly modeled. What if the conventional wisdom about the value of such tools is in error?
But this just recasts the problem. If the lesson is that despite pollsters’ best efforts, the data is too coarse and too difficult to model to produce useful predictions, then how is it helpful? How does a range that wide, based on a sample that may or may not be reliable, help allocate a campaign’s resources?
And, even worse, a large fraction of readers view simple, declarative statements as “honest,” while they view complex statements (especially if they describe ambiguity!) as “dishonest.”
I haven’t seen this combination of facts put together in any of the post-mortem analyses: Did the premature acclimation of Hillary lull some of her potential voters into non-voting complacency? “She’ll win, so I don’t have to turn out.”
If that’s so, that’s possibly the saddest indictment of the current political ubersystem that I’ve ever heard.
The pols were not wrong. They said Hillary would get more votes than Trump, and they were right. What the pollsters could not have known was how many likely voters would just decide on election day not to vote, which resulted in a 10-millon voter shortfall from what would follow the trend of recent elections. Disgust enough to not vote ran deeper among Trump-haters than among Hillary-haters.
One thing worth keeping in mind is that this election is a failure of public polling data and (especially) analysis.
Campaigns conduct their own internal polls and analyses that they don’t generally share (except for strategic reasons).
I would guess that once those start dripping out (and maybe they have - I’m being selective in my media diet, to say the least!), we may find out that there was a lot of uncertainty that the campaigns knew about but the public did not.
PS: loved dracol’s explanation of statistical uncertainty. That gets left out of news analysis too often. Or reported clumsily. One reason I encourage high school kids to take stats instead of calculus
It’s a theory that I’ve heard from several IRL people - I would imagine it is out there in the mediascape.
Testing it will require more … data, though.
PS: I am skeptical that it explains the loss, because people tuned in enough to know about vote projections would - I think - be politically committed enough to vote anyway. By and large. But I could be wrong.
I think that’s part of it, combined with the fact that many people weren’t that enthusiastic about her anyway and many weren’t happy about the whole election fiasco either (two unlikeable candidates). A perfect storm of apathy.
Internal polling wasn’t wrong, because the candidates campaigned where they needed to. The public polls are a type of propaganda, at least. They are designed to shape public opinion to some degree, not gauge it. Any idiot noticed crowds of tens of thousands for Trump, while the Clintons couldn’t draw flies. The momentum was never there. Bill showed up here in my hometown and the newspaper said a crowd of 300 showed up. It looked like maybe 100 people, including the press, and the photos carefully framed to make the crowd as large as possible. At the time I thought that was not a good sign for the Democratic candidate. It is a strongly democratic town, for the most part.
The good thing about the WikiLeaks and similar revelations have shown the corruption and collusion of the large media outlets with the democratic party was far beyond what most had imagined. The media has lost what little credibility they had. Trust once lost is very difficult, if not impossible to regain. Before the night was over the talking heads were reduced to casting aspersions on the electorate in flyover Jesusland. They still don’t get it. Journalism is dead. Good riddance to bad cess.
There are some really good replies in this thread, a reminder of why I read the Dope. I almost feel underqualified to join in, because I’m not certain I can adequately explain my point…but sometimes you have to take into account anecdotal evidence. I live in a very conservative area. Republicans always carry the elections here. I knew there were a lot of people who simply would not vote Democrat, and certainly not for a woman. There was a noticeable lack of Trump signs–I have only seen 2 actually in someone’s yard here in town–and none at all for Clinton. The shortage of Trump signs gave me a very faint hope, but I knew overall that he would carry the area. So, while there was a lot of coverage of people expressing disgust over Trump and wondering how any reasonable person could actually vote for him, I knew that there are a lot of forgotten flyover areas where people would do just that, and I wasn’t fully confident, not as confident as I would have been if I lived in a more liberal area.
The OP mentioned the housing crash. That was another time when I remember being uneasy, because here I was in this little town where the economy is based on agriculture and where a significant portion of the school kids qualify for subsidized lunches, and people were building these massive houses that I knew they couldn’t possibly afford. I knew these people. There was no way they were making that kind of money with a high school education. Because I’m not very skilled in economic matters, I did not foresee the impact the crash would have nationwide, but I did think, “Oh wow, these people are nuts.”
There’s what you see in the media and there’s what you see looking around yourself, and you have to try to put it all together, and sometimes it’s like being in a small earthquake and wondering when the ground will stop shaking so you can think.
Astro I like that you mention the housing crisis.
I lived in the US from 2003 until 2006, and throughout that period it was obvious that the US was in a housing bubble. At that same time millions of people refused to accept that this was happening, believed that real estate prices would never go down.
There were warning signs but they were ignored by the hopeful.
They were also ignored by – as you said – the confident. Confidence is seductive. It’s also, when it comes to predictions, divorced from reality.
I think a similar thing happened with this election.
Polls were not wrong, but the analyzers provided the wrong interpretation.
For one, national polls are of very little value when it comes to choosing the US president. Because of the Electoral College, what matters more is the value of support in each state.
Getting accurate impressions of a large group from a sample is fairly straightforward, which is why people conduct national polls – but when you want to break it down to smaller groups your margin of error grows larger. So, really, you need to conduct a whole bunch of state polls and then discuss them.
If your model is bad, good data won’t help. If the data is bad, you’re sunk.
I followed the polls closely before the election, and they were broken down by state. Even if Clinton won the popular vote, I don’t know how you could conclude the polling and predictions were not a complete shambles; margin of error is important but the differences in Ohio or Wisconsin surprised me.
There is a real echo chamber effect. All the journalists I know hated Trump – they resented being called luegenpresse or being treated with extreme disrespect. I think there was overcompensation both ways – trying to strike balance by reporting on Clinton’s emails and not more positive things; making blanket statements about how women or other groups felt about Trump that made sense but were obviously not even close to universal.
Canada has a radio station called CBC. It has some good news and programs. First Nation aboriginals in Canada have been treated unfairly, and I think it’s important to acknowledge past errors and inequities and provide compensation. But it seems recently that every time I turn on CBC, this is the only topic of conversation. I just don’t want to hear about this all the time, even though it has value and even though it is important. I struggle to understand American attitudes to race and immigration and wonder if there are similarities – the Democrats are so inclusive and politically correct that it can be too much even for some sympathizers, and of course not everyone is sympathetic. I apologize if I have stated this badly.
Of course, Trump supporters had their own echo chambers and many in small towns allegedly did not know anyone voting for Clinton. One reporter counted the number of signs they saw on a long road trip, and there were vastly more Trump signs, of bigger size, which some supporters had enthusiastically customized. There were only two or three Clinton signs. So the data was there. The reporter thought it was an isolated pocket.
I think it is commendable the Democrats see themselves as the party of women, Hispanics, LBGTQ, blacks, etc. and support that. This should be only one of many things that characterize a political party. Perhaps it was overplayed, to my surprise a majority of white women voted for Trump. I didn’t see this at all in preelection polls, and don’t think it was close to margins of error or that so much changed in a couple days.
