Lies, Damned Lies and Election Statistics

I was fair to Michelangelo in my Sistine Chapel post, but misrepresented the estimable Sam Wang.
I wrote:

Sam Wang gives Hillary Clinton a poll-based 70% probability  of becoming US President, if the election were held tomorrow ..

That was wrong and I apologise. Wang’s poll-based estimate if the election were held tomorrow is 99%. The 70% is a forward-looking estimate for November, based on the polls to date plus one other piece of information: the historical variance between polling in May and results in November over the last 60 years.

This is informative, and Wang’s austere poll-based purism is a useful corrective to horse-race and economic-fundamentalist narratives. But it’s far from compelling. Wang himself writes:

But considering the upheaval in the Republican Party, a little voice tells me to open my mind to a wider range of possibilities … including a Trump win.

There is no deep reason why we should not peer behind the curtain of historical variation and use the non-polling information we have to improve on what it suggests. We skirt the impenetrable mato of statistical philosophy Brad deLong ventures into here; I hope we see him again. As normal gamblers, let’s take as the question: what reasonably foreseeable factors are capable of changing voters’ preferences for Trump or Clinton between now and the election? I score the downside risk factors unscientifically out of 100, as a prop.

1. Public characters and policy positions (or attitudes in the case of Trump): these are very well known, including Trump’s propensity to change his policies from day to day. Change risk: nil.

2. Skeletons in the closet. Clinton has been the target of more oppo research over 30 years than Jack the Ripper, with trivial results. Risk 5. Trump’s picaresque personal and business past has been far less investigated, and at first sight is full of problems. Risk 40.

3. Gaffes. The calculation and lack of spontaneity that make Clinton unattractive to many armour her comprehensively against gaffes. Her performance in the Benghazi! hearings was masterly. Risk 2. Trump’s whole shtick is to make media-worthy gaffes and then withdraw them. It appeals to his base, but alienates constituencies he needs to get elected. His casual floating of the idea of renegotiating the US national debt – that is defaulting on it – must have lost him much of Wall Street, and hundreds of millions in potential donations. The Great Wall fantasy has guaranteed total support of Clinton by Mexican-American Latinos, and may have cost him Florida by Cuban-American sympathy. We can confidently expect more of the same. Risk 90.

4. GOTV. By this I mean more than the mechanics of a ground organization, but the entire process by which a candidate crowned by a party’s activist base is recognized as okay and voteworthy by non-activist members of each tribe. Clinton has a proven campaign machine, having ditched Mark Penn and defeated a serious challenger in Sanders. She will inherit Obama’s even more effective GOTV machine; Trump will fill any deficit in enthusiasm. The centre-left policies that have caused her problems with Democratic activists are, I think, ideally matched to the non-activists, not by accident. Could Sanders’ supporters stay away on a much larger scale than supporters of previous losers in Democratic nomination contests? Possible, but I can’t see why. In contrast, Trump barely has a ground organization, his nomination campaign having been based on media manipulation. Worse, many of the positions that have roused his base are deeply unattractive to non-activist Republicans. Making the necessary bond with them will be an uphill struggle for him, maybe impossible. Risks to Clinton 5, to Trump 75.

5. Money. No change for Clinton, but it’s an issue for Trump. In Wang’s historical dataset, the Republican candidates were all able to appeal to wealthy conservatives for donations. The Kochs at least are signalling that they will sit out the presidential election and put their money into Congressional races, with the far more realistic objective of keeping a Republican House and Senate. A wider plutocrat funding strike – not certain but a real possibility – would damage Trump. Could he fill the gap with his own money? Possibly, but at some point self-preservation would win over vanity. Risk to Clinton 0, to Trump 40.

6. Media narrative. The MSM require a close race for eyeballs, and will do their best to manufacture one from straw if need be. However, if Clinton keeps a commanding lead into September say, the commentators may throw in the towel, as with late-season matches between sporting teams out of contention for the championship. In that case Trump’s tweeted provocations may not be enough to keep him in the news. The risk to the Democrats in this scenario is not to Clinton herself, but to down-ticket races where low turnout could cost congressional seats. Risk to Clinton 0, to Trump 25.

7. Assassination by a lunatic. Same low risk for both, say 2. Mainly important in the public assessment of VP picks.

8. VP pick. The CW is that a good pick adds little to a candidate’s chances, but a bad one (Palin, Eagleton, Ferraro) can lose votes. The bar is low, see Quayle. Clinton will pick somebody sane, competent, bland, and vetted, better and probably much better than Quayle. Risk 0. Anybody who Trump asks to run with him, and accepts, will be missing at least one of these characteristics, perhaps spectacularly. He asked Ben Carson to lead the search. Risk 30.

9. The economy, stupid. The current recovery in the US economy is anaemic but steady. Wages are beginning to rise. There is no inflationary spike that could lead the Fed to kill the recovery. More of the same benefits Clinton. The risks are international: the Chinese slowdown may be worsening. Trump would benefit from a crisis severe enough to halt US growth. However a full-blown international panic would benefit Clinton as a safe pair of hands. Risks 15 to Clinton, 50 to Trump.

10. Terrorist attack. To change the odds significantly, this would need to be large and on American soil. An atrocity in Syria (more video’d executions of American hostages) would not IMHO do this, lacking novelty. There is no evidence that al-Qaeda or ISIS are at present capable of mounting such an attack. ISIS’ toy caliphate is shrinking slowly on the ground, it has problems paying its fighters, and regularly loses leaders to air strikes. If there were a big attack, it would harm Clinton. Trump’s lack of any coherent strategy would not count against him in the face of a visceral patriotic reaction. Risk factor to her 10, Trump 0.

11. Black swan event. I said “reasonably foreseeable.” Clinton may have a complete breakdown during a televised debate and be carted off to mental hospital under sedation. Trump may start plagiarising Paul Krugman’s columns and policy papers from the IMF, or admit he’s in the race for the exposure and doesn’t really want the demanding and (by his standards) badly paid job. Rupert Murdoch and/or Roger Ailes may have incapacitating heart attacks. The supernova shock wave that’s been heading our way for a few centuries may arrive and flash-boil us all like lobsters as the oceans turn to superheated steam.

Adding these pseudo-numbers up, I get the total risks to Clinton 39, to Trump 352. Really the only more than marginally possible future events in my categories that he has going for him are ISIS pulling off a big atrocity and economic collapse in China, both at long odds. I don’t claim credibility for my particular numbers, just that overall we have to put a very fat thumb on the probability scales in Clinton’s favour. So her chances to a sensible bettor are more than Wang’s 70%, a lot more.

My own unscientific hunch is that the presidential election will turn boring in September. Clinton’s lead following the conventions will go firm at a roughly 10% margin and the political class will come to accept that she is sure to win. Their attention will shift to the congressional and state races, of which some are bound to be close enough for aficionados to buy tickets.

Now you may comment on the election, dear readers. What factors in Trump’s favour have I missed out? To keep some focus in the thread, let us stipulate that the candidates will be Trump and Clinton. And please try to stick to the question: what could change voters’ minds between now and November? This means ignoring the information (such as party identification, propensity to vote, and views on Obama’s record) that is already baked into current polls.

[Cross-posted at The Reality-Based Community]