I’ve found Scott Adams to be the most prescient commentator I’ve read on the major issue of the campaign up to this point – Donald Trump.
Here Adams lists some of his calls.
Since it’s my job to model elections, and since Trump’s tactics have flummoxed most political experts, it’s obvious that the experts have things wrong. My guess is that Adams is correct that Trump is something like a ‘Master Persuader’.
The question for me is how to add this to the game. I’ve increased his Charisma to 4, and his Spin to 4. On the to-do list is increasing the affects of spinning and interviews in a campaign, since up to now that’s how Trump has operated. I’m also considering implementing a ‘contentious’ variable for speeches and so on, since Trump’s modus operandi seems to be to provoke by ‘speaking in headlines’, as someone put it.
I’m going to look carefully in the next few days at Adams’ analysis, and see if there are other elements I can apply to the game mechanics. What do you all think of his hypotheses?
Similarly, were there other writers who have had a similar track record in predicting what was going to happen in this campaign?
(On a side-note, I am amazed at how little people care about political pundits’ track records. Most get things wrong again and again, yet many people keep listening to their predictions and giving them weight.)
If you’re going to look at him, I’d suggest equally looking at Nate Silver.
Thanks for this – I’ve also looked at some of Silver’s stuff. In particular, I found his attempt to retrodict Republican poll numbers by a model combining net favorability and share of media coverage interesting.
http://fivethirtyeight.com/features/trump-boom-or-trump-bubble/
I’d say Nate Silver is almost never wrong. Harry Enten, who also writes on his cite, is less reliable. He sort of operates more on gut instinct, which is always interesting, sometimes correct, but often kind of quixotic. For example, and it is still early, he has some strong fascination with Chris Christie ultimately becoming the establishment choice against Ted Cruz or Donald Trump.
Silver has made at least a couple recent major errors in his predictions.
The first was the recent British election.
http://fivethirtyeight.com/datalab/what-we-got-wrong-in-our-2015-uk-general-election-model/
More relevantly here, he thought Trump’s support wouldn’t last ‘in the long run’. Writing in July, he said
“In the long run — as our experience with past trolls shows — Trump’s support will probably fade. Or at least, given his high unfavorable ratings, it will plateau, and other candidates will surpass him as the rest of the field consolidates.”
http://fivethirtyeight.com/features/donald-trump-is-the-worlds-greatest-troll/
It sounds pretty straightforwardly wrong, but perhaps one could be generous with the term ‘long run’. (> 5 months since the prediction was made.) Trump’s support has not faded or plateaued, it has grown since the time he wrote that. The rest of the field hasn’t consolidated much either. Rather, the rest of the Republican field has taken to attacking each other instead of Trump.
@Anthony – 5 months isn’t the long run, especially in a contest like this. I would wait until after the primaries to make any conclusions like that.
As for Silver, he’s one of the few people that analyze why they got it wrong. I think he’s a better source for predictions than the hypnotist.
@Nick and Anthony
Yeah, I’d say Nick Silver has way more credibility than a hypnotist as well.
@Nick,
In the context Silver was talking in (‘troll’ candidates), 5 months seems like a long run. Sure, he might have meant ‘Trump won’t win the nomination’, in which case we couldn’t evaluate it until March or April (or later), but that seems like a different claim. He could have easily said so (by March or April of next year, Trump will no longer be in the lead), but he didn’t, settling on a vague term suggestive, again, in the given context of a few months, it seems to me.
If you think he’s a better source for understanding the Republican primaries than Scott Adams, then why has Adams been consistently correct so far in his predictions, ones that Silver failed to make? Has Adams just been lucky, or is he understanding something about the Republican primaries that Silver – for all his data-obsessed wonkishness – doesn’t?
Well, Scott Adams is highlighting his own triumphs on his own blogs so I’d take it with many grain of salts. It’s like reading the Republican Party (or Democratic) platform and going “Jeez, these guys really know what they’re talking about.” The list you linked to all seemed relatively vague. His guesses are more vague than mainstream predictions. In at least a few cases, Adams is simply reading what the mainstream media is saying (Fiorina and Carson topping out).
1. Trump would gain popularity and win the nomination, not fizzle as every other pundit predicted. (I predicted it in August, based on his skill set. So far, so good.)
Uh, not so far so good. There have been no primaries. He is leading in the polls, yes. That’s it.
2. The Jeb Bush “low energy” kill shot would end Bush. (First to predict it.)
This assumes the outcome of the Republican primary. It’s also vague. What IS “killing” Bush? He’s still in the race, he is polling lower — that is true.
3. The Fiorina top in her poll numbers (after she paired her own image with a dead baby)
Summarizes what the media was saying on TV already.
4. The Carson top in his poll numbers (after Trump did his famous belt-buckle speech)
Summarizes what the media was saying on TV already.
5. The Clinton top in her poll numbers (after Trump noted how many women her policies have allegedly killed)
What? I don’t even know what this is referring to or when. Clinton’s primary poll numbers jumped once Biden “exited” the race, and they’ve been relatively consistent for the general from what I’ve seen.
6. Trump’s “nice guy” move that involved going into a crowd to personally help a wounded warrior with the Veteran’s system.
Also, what is this referring to? This is also vague. “One day, Donald Trump will be a nice guy.” Wow.
7. You will start to see Freudian slips in the media calling Trump “President Trump.” And so we have.
People have also said President Clinton…President Cruz….many reporters phrase questions as “What would President XXXX do in this scenario?” Obama even referred to Biden as “President” at least once, it’s like guessing that someone will eventually trip. Come on.
8. My 3D predictions – no matter how accurate – will be ignored by the standard 2D media. Check!
Don’t know what this is referring to. Also, can you ignore someone if you don’t know who they are? This is like me saying some guy in China is ignoring me.
9. [Update] My prediction months ago that Trump’s persuasion skills would set off a swarm of competing (and wrong) explanations for why Trump is defying expectations.
Oh, and #9 — I do think he is relatively spot on about Trump’s persuasion skills for the most part. The right/wrong explanations for why “X” candidate is doing “good/bad” is vague…again. Everyone talks about why “X” candidate is doing “good/bad” and some are inherently correct or incorrect.
@Nick, thanks for this feedback – noted.
As for 8., Adams is fairly well known (creator of a highly popular comic strip).
This is one of the biggest mistakes when people criticize FiveThirtyEight: They say Silver’s predictions have failed, when in reality they are still playing out. Silver (and Harry Enten) have both said they’re forecasting the probability of winning the nomination.
They say they Trump can poll in the lead for ages, and win some primaries, but that has to be distinguished from the likelihood of winning the nomination (due to the fact that ONCE STATES START VOTING – the long run – candidates drop-out and consolidate).
http://fivethirtyeight.com/features/polls-suggest-trump-will-win-between-8-percent-and-64-percent-of-the-vote/
“[It’s false that] Trump’s lead is somehow more meaningful because it’s lasted this long. Winning a pretty meaningless metric for a long time doesn’t magically make that metric meaningful.”
So I think it’s absolutely wrong to say they’ve made “a couple of recent major errors” about Trump for two reasons. A) They acknowledge that he could win (improbable but not impossible because it would be unprecedented). B) The field needs more time to consolidate post-Iowa and NH. http://fivethirtyeight.com/features/donald-trumps-six-stages-of-doom/
@Stephen,
In this specific case, I said that – under a generous interpretation of ‘long-term’ given the context of the article from July, where the prediction I’m referring to was made – one could say things are still playing out.
Yet, that seems like a *generous* interpretation, given the original article (not ever-changing predictions as things play out).
“So I think it’s absolutely wrong to say they’ve made “a couple of recent major errors” about Trump for two reasons. A) They acknowledge that he could win (improbable but not impossible because it would be unprecedented).”
It sounds like you’re saying they wouldn’t be wrong even if Trump wins the nomination, because they said it’s not impossible! Sorry, that’s not how it works. If Trump wins the nomination, Silver’s original prediction I am referring to will be definitively shown to be incorrect (generous interpretation of ‘long-term’ or not).
Regardless, I would say that their initial assessment of Trump has been wrong, and that the generous interpretation is the incorrect one. For example, here’s Silver in his most recent article, saying
“If you, like us here at FiveThirtyEight, were initially skeptical of Donald Trump’s chances of winning the GOP nomination in part because you assumed that the Republican Party would go out of its way to stop him, then you’ll find the following pretty remarkable. According to Tim Alberta of the National Review, there are currently no negative television ads running against Trump in Iowa, New Hampshire or South Carolina.”
http://fivethirtyeight.com/features/why-some-gop-candidates-arent-taking-the-fight-to-trump/
Note the word ‘initially’. Silver is saying he thought the Republican Party would go out of its way to stop Trump, and that assessment was the basis of his belief that Trump’s chances were low. Yet, it hasn’t yet done that. Therefore, the basis of his belief has been undermined. He’s basically paving the way for explaining why he was wrong, if it happens that Trump does win the nomination.
(Forgive the length). 5-months is not a generous interpretation of long-term; the only fair reading of long-term is ‘by the end of the nomination process’ since these predictions refer to a process that won’t finish for more months still (i.e. the convention). You’re misreading the July quote, “In the long run — as our experience with past trolls shows — Trump’s support will probably fade. Or at least, given his high unfavorable ratings, it will plateau, and other candidates will surpass him as the rest of the field consolidates.” (see ‘6 stages of doom’ where he spells it out e.g. the consolidation is “Stage 4: Winnowing, When it happens: ! mid-February through mid-March. !”)
You’re right that Silver admits they underestimated Trump’s staying power initially, but they have not retracted their overall low-rating of Trump’s chance of winning because the basis of their predictions was not simply that the Republican party would fight but an array of scientifically meaningful variables (based on historical statistics).
Even if Trump wins, I am indeed saying Silver will have been accurate. Why? Because 538 (unlike the vast majority of pundits) is very clear to outline uncertainty and has said that improbable does not mean impossible. Evidence for why Silver is a more better forecaster to read over the long-term:
1/3) http://fivethirtyeight.com/features/how-the-fivethirtyeight-senate-forecast-model-works/
“A good model should be probabilistic, not deterministic… probabilistic forecasts can be very sensitive. Does that 3-point lead translate into a 60 percent chance of winning? Or a 95 percent chance? The best test of a probabilistic forecast is whether it’s well calibrated. By that I mean: Out of all FiveThirtyEight forecasts that give candidates about a 75 percent shot of winning, do the candidates in fact win about 75 percent of the time over the long run? It’s a problem if these candidates win only 55 percent of the time. But from a statistical standpoint, it’s just as much of a problem if they win 95 percent of the time.”
So since he lays out the evidence why Trump is unlikely to win, but if you could run multiple simulations of the universe and Trump won none, his estimation would be too high; Trump should win in 5-10% of simulations, according to Silver. Also, we should thus expect that percentage to change the more stages of the primary he goes through: predictions evolve over time i.e. Cruz may have a 25% chance of winning the nomination now, but by July his chance will approach either 100% or 0% (as he either locks up the nomination or drops out). But the early prediction of 25% will (hopefully) still have been correct; same goes for Trump: if he realizes the 5-15% chance by winning, it won’t make that initial estimate wrong since it’s derived from the objective data.
2/3) http://fivethirtyeight.com/features/despite-keys-obama-is-no-lock/
“If two interns applied to FiveThirtyEight, and one of them claimed to have a formula that predicted 33 of the last 38 elections correctly, and the other one said they had gotten all 38 right, I’d hire the first one without giving it a second thought — it’s far more likely that she understood the limitations of empirical and statistical analysis.”
3/3) He usually gives Trump 5-15% chance of winning nomination (pro: polling, con: high unfavorability ratings, low ‘second-choice’ in polls, a win would be unprecedent but there is a low sample size, and no endorsements all suggest a factional candidate who will be eclipsed as the field consolidates post-Iowa/NH). The reason he gives 5-15% at this stage and not 0% or 100% is because he is following the evidence: the low-sample size and Trump’s lead are not meaningless; just that other more historically reliable indicators (endorsements, IO-NH vs national polling disparity, no other factional candidate like Buchanan Huckabee winning) point in the other direction.
This article is probably the best synthesis, which is only from Jan 8th: http://fivethirtyeight.com/features/three-theories-of-donald-trumps-rise/ “National polls, even with barely more than three weeks to go until the Iowa caucuses, aren’t highly predictive of the eventual outcome of nomination races… And even if you take them with a dose of skepticism, could he win the nomination anyway? (Spoiler alert: Yes, although I continue to think Trump’s chances are lower than where betting markets put them – I think they’re about 12 or 13 percent.”)” He lays out the different theories, notes which he thinks is more accurate, and how Trump could still win even if the theory is accurate.
Silver is trying to put an objectively data-based probability to this event, whereas most pundits are saying ‘x/y is likely to win.’ I trust Silver more because he’s clear about what he means by probability, clear about uncertainty, honest when he thinks he over or underestimated, but each and every time he lays out the statistical data to support his reasoning – and all that is sadly uncommon in political prediction. I’d make the analogy with sports: if you say the best team has an 80% chance of winning their games, you’d still expect them to lose some games, and even lose some series, just as the worst team with a 10% chance should win some games, and once in a while a bad team should have an unlikely series win; even while the underlying data and theory is sound. A Trump win by itself wouldn’t cause me to doubt Silver, but if Trump-like candidates were to win the next several nominations and he stuck by the endorsements-is-stronger-than-polls model, then I’d reconsider.
Last addendum; I’ll rephrase why I think Silver is the most credible forecaster: The problem with political prediction is that test cases happen so infrequently (every 4 years, and only about 100 years with something resembling full enfranchisement, and 50 since the Presidential primaries were made democratic). This creates a small sample size and it makes it hard to test. So having a single failure (Trump loses/wins) will invalidate the many overconfident pundits who said x will win/lose (with a high degree of certainty); but since Silver is more accurately data-based, the real way to test would to gather estimates from, say, each cycle (e.g. 2015, 2011, 2007, 2003, 1999, etc.) and check: did the candidates with a 10% chance of winning the nomination win the nomination about 1 in 10 times and lose 90%; did the favorite (say 80%) win 80% of the time. The only way to test that is with multiple ‘runs.’ To his credit (and conversely to the fault of pundits) Silver and 538 are producing models that can actually be tested in multiple elections and put their money where the mouth is since you can test their accuracy.
“Even if Trump wins, I am indeed saying Silver will have been accurate. Why? Because 538 (unlike the vast majority of pundits) is very clear to outline uncertainty and has said that improbable does not mean impossible. Evidence for why Silver is a more better forecaster to read over the long-term:”
Aye yay yay!
So heads, Silver wins. Tails, he wins. Sounds kind of silly to me.
Yes, 538 often assigns probabilities to their forecasts, but almost every pundit does that (in more or less explicit language).
The question I have is whether in this case they forecast it correctly.
One more thing, and then you can have the last word if you’d like it.
Political elections aren’t the same as random events – we are not talking about a possible quantum randomness effect here. Elections are the result of planning, abilities, instincts, strategies, tactics, infrastructure, and so on.
Assigning a probability is a reflection of the forecaster’s limited knowledge of the various elements, not a reflection of objective reality. That is, the forecaster isn’t saying that if you ‘re-wound’ time to the beginning of an election, and then let things play out, 15 times out of 20 (or whatever) party A would win.
If that *is* what Silver and co. mean, then their hypotheses are untestable, because we never have enough similar elections like this. In other words, if that’s what they mean, they are just blowing smoke.
If, on the other hand, they mean that with their limited understanding of the variables, they can say with 95% confidence that so-and-so will win (‘very confident’), then if that person doesn’t win it shows that their model is incorrect.
“So heads, Silver wins. Tails, he wins.” Not quite – I’m saying it would take more than one simulation where a low-probability event comes true to disprove Silver’s model. Having Trump win is not sufficient – by analogy, imagine that you have a lottery where one person has 80% of the tickets and the other has 20% of the tickets; you can with confidence say the former is most likely to win. What you are saying is that if the 20% chance person wins, the model is incorrect; but objectively that’s false: things with a 20% chance to happen should happen 20% of the time, and the 80/20 forecast was accurate. With a lottery, you know the variables affecting the outcomes (number of tickets), and you can rerun it several times but you can’t do that with elections. So we have to identify the relative importance of variables (endorsements, polls, etc.) that you can test across multiple elections. 538 seems to be the only mass media taking this scientific approach: “The best test of a probabilistic forecast is whether it’s well calibrated. By that I mean: Out of all FiveThirtyEight forecasts that give candidates about a 75 percent shot of winning, do the candidates in fact win about 75 percent of the time over the long run?”
Silver’s model is the most sound out there (theoretical and historically – both with how it fits past elections, and his previous relatively accurate predictions of 2008, 2012 President, and 2008, ’10, ’12, ’14 Senate races). You’re right that we can’t rewind time, but we can use the explanatory variables in a concrete, quantifiable method that can be applied to fit past elections and to see how well it does in subsequent elections (which 538 does, and practically no others outside of political science academics do). So because their model is falsifiable (all their math, and historical statistics are made available – i.e. you don’t just get the overall judgment, like “almost every pundit,” you get a model that is testable, and able to be verified over cycles), I give it the most credence.
(If you’re keen on leaving the conversation there – I do want to add I still love your game! =) ).
Does anybody still believe that Romney could get into the race? I wonder what would happen if Trump wins in Iowa, and it seems very likely according to the last polls and a good debate performance. If it happens, the GOP establishment would look for the best guy to unite the moderate voters against Trump. Romney is actually the best qualified for this job. Jeb is not charismatic enough, Christie is a joke, Rubio is a child and Kasich doesn’t have the national background to lead the voters. Mitt in the race before New Hampshire primary?
Here’s why Silver is most reliable: http://fivethirtyeight.com/features/the-republican-party-may-be-failing/ He is approaching politics scientifically, not ad hoc. When the evidence changes, he changes his views, but is very clear about the data supporting them. Incidentally, he also points to evidence illustrating his logic with either outcome; the Party could decide against Trump (for Rubio/Bush) OR decide FOR Trump (willing to throw a general election) to block Cruz. That would still be vindication for Silver (the Theory he subscribes to is ‘the Party decides’ not ‘x/y will win without supporting evidence’).