Predicting the 2020 Democratic Primary is Stupid | Pt 2. Correlation is not Causation
Since 1980, the Iowa Caucus has correctly predicted the eventual Democratic Nominee every time except once, in 1992.
Of those Democratic Nominee predictions, 3 won the popular the vote (Al Gore, Barack Obama and Hillary Clinton). The rest were crushed in electoral landslides (Walter Mondale, Mike Dukakis, John Kerry).
Of the 3 who won the popular vote, only 1 became President, Barrack Obama.
The year that Iowa incorrectly predicted the nominee, 1992, produced a President who had a plurality of the popular vote, but who still became President — Bill Clinton.
So, since 1980, Iowa has correctly predicted almost every Democratic Nominee, but has only correctly predicted the President once. What does all of this mean?
It means absolutely nothing. Zero. Correlation isn’t causation.
Turnout Determines Elections
Sometimes things can look like patterns that aren’t there. Your judgements about people’s behaviors can be off by an inch or a mile, depending on how you group them. Your ability to understand data about people’s behavior depends a lot on how well you group people, and how well you can guess about the behavior of the people you grouped together. It’s imprecise and messy, and in politics, no one is really doing much of that kind of work. A lot of political pundits create the illusion that they’re looking at data, when most of them are just riffing off the top of their heads on TV as a guest, and look smart while doing it.
I do know one thing about political data, though, the one that (to me) feels like it gets ignored the most.
When you have massive turnout, Democrats win. Period.
When Presidential season comes around, pundits use terms like ‘centrist’ or ‘swing voter’ and it all sounds very official. The pundits end up on CNN and Morning Joe or The View, and it all sounds so logical. So, then the rest of us use the same terms, and we argue, and that’s all fine, except those arguments influence how people behave. They change how people talk at caucuses. And I’d like to say, it’s all bunk. There’s very little data that supports the terminology used, that supports most of what our political media talks about. It’s why guys like Chuck Todd have a job; they can speak the conventional wisdom in a way that hides that nobody knows what they’re talking about.
Here’s a question nobody is asking me — why did I feel so confident in the blue wave of 2018–2019? It’s because I knew that the turnout rates were higher than average that year. Here’s a recent story about the correlation between turn-out and Democratic wins, something a lot of savvy organizers and data scientists have known for decades.
For the record, the percentage turn-out in 2018 was 49.6% of the population. That’s the highest ‘mid-term’ election turnout in an election in American history. The most people in the history of the country voted in a ‘mid-term’ in 2018.
So, here’s the deal. Anyone can win the 2020 Presidential election. Anyone. What’s going to determine the winner is turnout.
And my big question, the thing no one is talking about, is what does it mean that turnout for the caucus in Iowa was down this year? What does it mean that the New Hampshire primary did the opposite, with a record turnout? Can we trust the record turnout of New Hampshire, given Republicans have been vowing to switch parties to vote in the primaries?
If you want your candidate to win this year, if you’re passionate about it, if you want to beat Trump, then ignore pundit wisdom, and simply pour your efforts into turning out the vote. If 300,000 people hadn’t been disenfranchised in 2016, which is 00.12% of the total U.S. voting population, or 00.24% of the people who actually voted that year, Trump would’ve lost. You can argue about how or why those people in those key states didn’t vote, but the math is solid.
Register people to vote. Encourage people to vote. Nothing can defeat that. Not Russian bots, not Trump, not hacking. If enough people vote, Democrats win.
But who cares about all this preaching and civics talk? Let’s do something more fun, just like big media folks, and make a prediction about the upcoming Nevada Caucus based on a poorly conceived statistical model that you can read more about here!
Making A Bad Prediction About Nevada!
The Nevada Democratic Caucus is on February 22nd! There are 36 total delegates available. Based on our model, which assumes that the outcomes of Iowa and New Hampshire perfectly predict the future, we can assume that these are the delegate totals for each candidate in Nevada.
Buttigieg — 35.38% | +13
Sanders — 32.31% | +12
Warren — 12.31% | +4
Klobachar — 10.77% | +4
Biden — 9.23% | +3
El Bloombito — 0% | 0+0
Buttigieg will once again be neck and neck with Sanders, Warren and Klobachar will tie, Biden will be the furthest behind except for Michael “El Bloombito” Bloomberg, who will get 0 because nobody likes him.
That means none of the candidates will be on track to have enough delegates to win in the first round of voting at the 2020 Democratic Primary.
As noted last time, let me be 100% clear; I don’t believe that the Iowa or New Hampshire primaries are going to be that predictive of the eventual 2020 Democratic Nominee. I think multiple news outlets and pundits will make those claims, based on their historically bad metrics for predicting outcomes.
I do think, however, that not enough of those news pundits take their own work seriously enough in the right way. So, I’m going to keep finding 10 minute chunks at work to update this document at various intervals.
Let’s see how badly this predicts the future!
Very Bad 2020 Democratic Primary Projector
Thank you for reading this public civics announcement. Here’s another cat gif.