I haven’t posted here more than once since the debate, having been in Tevye mode since it happened. (“On the other hand…on the other hand…on the other hand…”) I’m still mostly on the fence, although I lean more towards wanting Biden to step down and Harris to head the ticket. However, one of the things that gives some pause is historian Allan Lichtman and his “Thirteen Keys To The White House” formula.
Lichtman has crafted a formula that appears to have been right for every election for the last forty years except 2000 (and many believe the SCOTUS helped steal that one and was an unforeseen wild card). In Lichtman’s view, if six or more keys turn up “false”, the candidate loses.
He’s been in the news recently for his adamant belief that it would be a mistake for Biden to withdraw from the race, since it would discard at least one key, the “incumbency” one. According to him, the only way to salvage that would be for Biden to resign the presidency completely, leaving Harris the incumbent.
I’m not quite sure how I feel about this. On the one hand, the Keys have correctly predicted every election for the past forty years save one. (He got quite a bit of attention when he was one of the only ones predicting Trump’s victory in 2016, even when polling said otherwise.) On the other hand (hello again, Tevye), I don’t know if incumbency is still the advantage Lichtman thinks it is, given the fact that quite a few voters (and insiders on Capitol Hill) have such serious concerns about Biden’s health and ability to pull it off.
So…what do you guys think? Do you put any stock in Lichtman and his keys?
I would question whether “incumbent advantage” even applies to a guy like Biden. I wouldn’t question it if we were talking Clinton 1996 or Obama 2012, but Biden isn’t the average incumbent. He’s 81, and also has very low approval ratings for an incumbent. If this were a 61-year old Biden, he might have incumbent advantage but it’s not. So right off the bat, I’d challenge Lichtman’s formula for not taking this variable into account.
Also, bear in mind that in the past half-century, incumbent presidents have lost just as many times (4) as they’ve won (4). Ford, Carter, Bush Sr. and Trump lost. Reagan, Clinton, Bush Jr. and Obama won. So a 4-4 record suggests that incumbent advantage doesn’t mean all that much anyway.
It’s a binary system that always predicts yes or no. That right there is enough to make it suspect. In an election like 2000, “Bush will win” is a bad call, and so is “Gore will win”. The correct prediction for an election like that is “too close to call”.
In fact, in any election, the output of a prediction technique should always be a percentage chance, not a definite yes or no. Maybe in some cases, that percentage chance will be in the high 90s. But it won’t always be.
I think politics is completely broken and past can no longer be considered precedent.
I mean, you’ve got a front-runner who is a convicted felon facing dozens more charges. A man proven to lie almost as much as he breathes. A man found liable for sexual assault and credibly accused of many more instances of assault or rape. A man who tried to overthrow the government to stay in power. He’s in the lead. He’s the anointed of the Christian right. He’s torn down the entire Republican party and remade it in fealty to himself, and they all cheer him for it.
Zero. Like all other such “predictive models”, the creator has successfully created something capable of predicting the past. Which, let’s face it, is much easier than predicting the future. And like all the others I’ve heard of over the years, I’m waiting for this one to fail this year so that I never have to hear of this particular hack ever again.
This has been nagging at me too a bit, having just heard about Lichtman recently. But I too think this year is too weird to fall into his formula. Much as it’s too weird for me to consider polling predictive in any way either. What was once helpful is now just entertaining nonsense, either exciting or terrifying depending on your POV.
Methinks that once this year is over he’ll rethink his Keys and add a footnote or 12.
Yes. Nate Silver argued this years ago. If you limit yourself to binary predictions, you are leaving data on the table. This might be acceptable if you had a large sample size. But that’s clearly not the case here: 100 years of data gives you a total of 25 observations, as well as the problems of shifting underlying parameters over time.
Predictions are a mugs game. Forecasts contain an estimated error term - or they are probabilistic.
Ok, say you have a forecast: you have an estimate and an estimate of the error. Your audience now wants a story to back it up. Lichtman’s checklist might be helpful at that point. So I don’t think it’s useless.
These “Keys” are arrant bullshit, most of them are just matters of opinion. The first six are objective (is the President an incumbent, is the GDP growing, etc). But the last seven are:
Major policy change: The incumbent administration effects major changes in national policy.
No social unrest: There is no sustained social unrest during the term.
No scandal: The incumbent administration is untainted by major scandal.
No foreign or military failure: The incumbent administration suffers no major failure in foreign or military affairs.
Major foreign or military success: The incumbent administration achieves a major success in foreign or military affairs.
Charismatic incumbent: The incumbent party candidate is charismatic or a national hero.
Uncharismatic challenger: The challenging party candidate is not charismatic or a national hero.
Pretty much by definition, there’s not going to be general agreement on most of these most years. Has the Biden administration been “untainted by major scandal”? Is the challenging party candidate a “charismatic national hero”? Depends on what network you watch.
Yet another reason to root for Biden dropping out and the Dems winning (which is where I was to begin with): Lichtman crawls back in his hole and is never heard from again.
People desperate to hear what they want to hear seem to glom onto these people and their “models”. It’s not really any different from the people who were for Romney listening to Dean Chambers “unskew the polls” in 2012.
I realize it’s subjective, but doesn’t Biden specifically now have a “scandal”. So losing incumbency but gaining “no scandal” is a wash (based on this faux-objective model).
I came here to write about Lichtman’s keys, and then Thing.Fish wrote what I was going to say, about a half hour earlier.
I’m not sure how Lichtman makes the call on the subjective elements of his keys, or if he’s fudged them before, but it’s not unlikely that he’s been progressively lucky with his calls on them for a while now. Past performance, as they say, is no guarantee of future results.
Specifically I do not believe the incumbency advantage exists right now. Across the world voters are wanting to change horses, from South Africa to France to the UK to Iran for gosh sake! Even in Mexico where the incumbent party won it was with a fresh face at the top.
You have Putin winning his election … but otherwise it is a bad time for incumbents.
A loved incumbent of course has an advantage, but I also generally see voters liking the change agent.
I think there is an opportunity to leverage that with the people not loving either of the two old farts.
I count seven true, one probably true but I’m not going to look it up, one probably false but INGTLIO, two false and one I honestly have no idea (does RFK count as a “significant” third party candidate?). So even if they punt on incumbency, the Dems should still be in good shape. YMMV significantly, however.
The flexible/vague metrics are there for a reason: they enable Lichtman to claim that he was never wrong.
If Biden wins, he’ll say he was right because Biden never had a scandal or social unrest. If Biden loses, though, he’ll say he still was right because he’ll then take some event from Biden’s presidency and call it a scandal or social unrest.
Thing.Fish. Everything is arrant bullshit. Polling is garbage. Nate Silver’s model is a steaming pile of adhoc assumptions built on a pile of garbage polling and coated with a thin veneer of bogus numerical pseudo-certainty intended to paper over the minuscule sample size. Pundits are even worse.
In other words, forecasting is hard and prediction is worse. There’s no good way to validate Nate Silver’s forecasts because of the thinness of the out of sample dataset and the necessity of large samples to make reasonable inference about the model. You say your model gives a 40% chance of something happening? You’re going to need a sample of over 20 to realistically figure out whether you’re really seeing a 50-50 situation. You say you’re 100% confident? It’s nonsense unless it’s validated by at least N=5 (.5^5 = .0325<.05. Ok, maybe N=4 is sufficient for barroom discussion. Still.)
So what to do? Up your science.
State your center-line prediction and your error bounds. Or state your prediction probabilistically. Nate does that.
Don’t throw away data. That’s where Lichtman’s approach comes in: he uses a checklist procedure. Yeah, some of them are subjective, but c’mon. And over-reliance of indicators that you are sure of tosses out information. Our sample is low so we can’t afford to do that.
That said, regarding the OP we’re in rough agreement. Lichtman’s work is billed as a great predictive tool and it really isn’t because there are no great predictive tools. There are only sensible or at least defensible forecast methods, something that Silver has done a great job of presenting to a wider audience.
Nitpick: Mexico’s presidents are limited to one (six-year) term.
(Or perhaps by “fresh” you were implying that Sheinbaum’s platform is markedly different from her same-party predecessor, AMLO).
Not important. I agree with the overall observation that people in much of the world are eager for change lately.