There have so far been 2 diaries hyping a 538 article written by Harry Enten that claims primary turnout does not predict general election outcome. Here is the 538 article: Primary Turnout Means Nothing For The General Election.
I left the following comment in both diaries, but I am having trouble locating the first diary posted by blue aardvark at lunch, about 6 hrs ago. Oddly, I can only find it by searching, blue’s profile or linking through Comments. I saw no rebuttals to 538’s conclusion in either diary and none on 538. I left a comment on 538 but it was deleted, probably because I added that it was sloppy work for a Dartmouth grad.
Here’s the comment. I was responding to someone’s claim that 538 showed that uselessness of Primary Turnout Margin is “fact”.
/Begin Comment:
lol fact. This is what happens when you don’t analyze statistics in context: Errors.
Let’s take a look at 538’s nonsense. Enten gives us the following table and says, paraphrased, “Look, 3 out of 6 and 4 out of 6 times. That’s hardly predictive.”
So, we see that for every year, except ‘80 and ‘88, the greater Primary Turnout Margin does indeed predict the Electoral College winner, a correct prediction 4 out of 6 times. So what happened in ‘80 and ‘88 that might have changed the predictive value of Primary Turnout Margin? What occurred during those election cycles that could have caused the Party with the larger primary turnout to lose the election?
During the entire 1980 election cycle we had more than 60 hostages, diplomats, Marines and staff blindfolded and paraded in front of cameras in Tehran. Mere months before the 1980 election, we had a failed rescue attempt that ended with the deaths of 8 servicemen and the destruction of 2 aircraft in Iran. Additionally, the Soviets had more than 100k troops in Afghanistan killing close to a million and destroying military equipment paid for and supplied by Jimmy Carter. Moreover, the US Ambassador to Afghanistan was kidnapped and killed a year before.
I’m not done yet. We had Stagflation (high UE Rate, high Inflation; no jobs, rising prices, pain) and a massive recession during 1980. I won’t bother looking up the economy’s numbers, but trust me, they were bad, very bad. And you could forget about buying gasoline for your car, unless you had a day off from work. The lines to buy gas looked like this:
Carter was also primaried by Kennedy. So, yes, even though Democrats had the greater primary turnout margin during the primary and still lost the general election, we can see plenty of reasons to explain the predictive error.
Ok, so what happened in 1988? Simple. An incumbent vice-president, who served during a greatly improved economy and an election landslide in ‘84, overcame the primary deficit to win the election. Not really hard to understand why the primary margin prediction failed there. We can also see this in 2000, with Al Gore, who was also an incumbent who served during a strong economy and relative peace. Although Gore suffered from a primary margin deficit, he was able to win the Popular Vote and he likely would have won the entire election if not for either Scalia or Nader. I know, many want to argue that Nader didn’t cause the loss in Florida. Sure, ok, I’ll grant you that even though I think it is false, but Gore definitely lost New Hampshire because of Nader which would have also given Gore the election outright. That really isn’t up for debate. Nader cost Gore the election.
So, by looking at the 538 statistics in the context of reality, we can see that it is rather easy to explain why primary turnout margin as a predictor of general election outcome was accurate only 4 out of 6 times, instead of 6 out of 6.
538 is becoming increasingly irrelevant.
This is what I wrote a week before Michigan primary:
While Nate Silver is very good at what he does, he is not infallible. Just ask the national soccer team of Brazil. Nate even started the primary season with a warning that primary polling is not all that accurate compared to general election polling, so all of his primary assessments should be taken with a grain of salt and a shot of vodka.
If Hillary is the nominee, which I very much doubt, we will witness the most gruesome political hatchet job election of all time, which we will have very little time to complain about before President Trump is sworn in and the USA ceases to exist as we know it. Trump pulled an amazing number of Dem votes in not just Mass but everywhere, and this is true before Trump even started swinging the the hatchet on Hillary’s past and present. It won’t be pretty and we will pay an enormous price if Hillary is the nominee.
Silver gave Hillary a 99% win rate.
lol statistical “facts” out of context.
/End Comment.
David Brooks apparently was on one of the Sunday shows peddling the same nonsense. But Brooks, foolishly, took it a step further. He stated that during the last 11 elections there was no correlation between Primary Turnout and General Election outcome. This is foolish because for 5 of those 11 elections there was a sitting president with no challenger in his party’s primary, which makes the low turnout number meaningless. You have to have a competitive election in each party to make the Primary Turnout Margin meaningful. Politifact is also pushing this and rated Brooks comment as True.
http://www.politifact.com/punditfact/statements/2016/mar/06/david-brooks/david-brooks-said-primary-turnout-doesnt-predict-g/
There will be no incumbents in this election. There, hopefully, won’t be any hostages abroad nor anywhere else. There is no recession and there won’t be for the rest of the year. No gas lines, and Russia is leaving Syria. Even if all these potential confounding factors did indeed exist, it wouldn’t matter in this election anyway, because neither candidate could be seen as responsible. Certainly not the candidate who already has the the turnout lead and the enthusiasm. Bush and Gore benefited from enthusiasm for their incumbent party, enthusiasm that allowed them to overcome the turnout deficit and win the general election.
Am I missing something? I really don’t see how someone experienced in statistics, as most people supposedly are at 538, could so blindly apply stats to complicated events and so easily reject the hypothesis. They also don’t even mention the extremely small dataset, just 6 years. But even that small dataset gives 4 out of 6 correct predictions, which Enten declares “hardly predictive”. When you put the 2 incorrect predictions in context of contemporaneous events, we are presented with a host of obvious confounders.
Makes no sense to me. I was thinking that it’s maybe Hillary partisans who want it to be true. I dunno. Can a Bernie supporter tell me, “yeah, it’s true?” You’ve got one Prof down in Florida also claiming turnout means nothing, no data or evidence, though. Others claim turnout is a factor. 538 is good in that they are the only ones putting numbers to it, but their analysis appears botched to me.
This is important imo. Even if you just look at the 4 out of 6 correct predictions, that would tell you there is a 66% chance Trump will win the election based on Primary Turnout Margin. I’ve probably forgotten more than I ever knew about statistics, so perhaps I am looking at that 66% probability incorrectly. I know it’s a simplistic way to look at it, I think. But if you remove the years with obvious confounders, then you have 100% accuracy. Trump wins.
Please show me where I’m wrong.
Finally got Twain’s pic up after 15 minutes of swearing at the computer and DK5. :P