Sunday, October 16, 2016


Remember the ridiculous election prediction I made on October 8, the day after the Access Hollywood tape was released, when women were just beginning to go public with their tales of being sexually assaulted by Donald Trump? I wrote:
I've been saying for weeks that it'll probably be a 2-point Clinton victory. Now maybe it'll be a 4-point Clinton victory.
Since then, more women have come forth with tales of sexual assault by Trump. His response has been to denounce his accusers as ugly liars before crowds who want the women locked up -- that is, when he's not making dark predictions of a rigged election.

So after all this, how big a blowout is he llkely to face in November, according to a new ABC/Washington Post poll?
... the race ... remains close, a testament to the strong pull of partisan preferences.

The contest stands at 47-43 percent, Clinton-Trump, among likely voters, with 5 percent support for Gary Johnson and 2 percent for the Green Party’s Jill Stein if the election were today.
Yup -- according to this poll, if the election were to be held today, it would be a 4-point Clinton victory.

Many Trump voters are backing him even though they think he did the things he's being accused of, and even though they think he's a bad person in general:
Trump issued a series of statements apologizing for what he said on the video, but nearly 6 in 10 likely voters say they do not think the apology was sincere, including more than one-fifth of Republicans....

... 30 percent of likely voters who support Trump say he doesn’t have a strong moral character.
Trump is a sexual assailant, and his voters don't care:
Two-thirds of registered voters say their preference is not affected by the release of an 11-year-old videotape in which Trump crudely described his sexual advances toward women. The one-third who say it makes them less apt to support him fit the profile of those who were unlikely to do so in the first place.
There will be more accusers, but it won't matter -- Trump voters are dug in. They hate Hillary Clinton and they'll tolerate anything from their candidate if they think he's the only one standing in the way of a Clinton presidency.

I know -- it's just one poll. In most other recent polls, Clinton has a much bigger lead. (A new NBC/Wall Street Journal poll shows Clinton with an 11-point lead.) But I think the impact of the sexual allegations crested a couple of days ago. It clearly left a mark, and it's certain to be affecting early voting. But by Election Day it could be background noise.


This poll doesn't ask about Trump's election-fraud conspiracy theories, but we know from a September ABC/Post poll that Trump voters think voter fraud is widespread:
Nearly half of Americans say that voter fraud occurs at least somewhat often according to a new Washington Post-ABC News poll, a viewpoint at odds with studies showing it rarely occurs in U.S. elections....

Over two-thirds of Trump voters say voter fraud occurs often, compared with less than one-third of Clinton supporters.
We've seen polling numbers like this:
In August, Public Policy Polling found 69 percent of Trump voters in North Carolina think Clinton would only win if the election was rigged: 40 percent actually blamed ACORN, which officially disbanded in 2010, as the reason they expected mischief.
There's a lot of concern about a Boston Globe article published yesterday in which Trump followers warn that they're not going to take it lightly if the voting doesn't go their way:
“If she’s in office, I hope we can start a coup. She should be in prison or shot. That’s how I feel about it,” Dan Bowman, a 50-year-old contractor, said of Hillary Clinton, the Democratic nominee. “We’re going to have a revolution and take them out of office if that’s what it takes. There’s going to be a lot of bloodshed. But that’s what it’s going to take.... I would do whatever I can for my country.”

... “Trump said to watch your precincts. I’m going to go, for sure,” said Steve Webb, a 61-year-old carpenter from Fairfield, Ohio.

“I’ll look for ... well, it’s called racial profiling. Mexicans. Syrians. People who can’t speak American,” he said. “I’m going to go right up behind them. I’ll do everything legally. I want to see if they are accountable. I’m not going to do anything illegal. I’m going to make them a little bit nervous.”
I think there's going to some intimidation on Election Day, and some violence after the polls close. That's bad enough. But I continue to suspect that Trump won't actually plot a coup intended to overturn the election. Doing that would be very different from running for president -- when you're running for president, there's an obvious series of steps you need to take (win these primaries, win these caucuses). In his shambolic, disorganized way, Trump was able to get through the primaries and run a general election campaign of sorts. If he wants to overturn the election results, I don't think he'll know what steps to take. I worry that Roger Stone and Steve Bannon, who just want to watch the world burn, might be able to lay out an agenda for him. But I don't know if he could manage the follow-through.

But the polls and reports like the Globe story suggest that Trump's voters will be ready if he does make the effort. They're angry. They think the vote is rigged. And they may be keeping the race closer than we thought (though I don't think they'll be any less skeptical about a landslide than about a closer Clinton win).

Other Republicans are going to be the key to all this. When outraged Trumpers start talking about fraud, what will they say? We're supposed to be heartened by Paul Ryan's words on this subject, but I have my doubts:
“Our democracy relies on confidence in election results, and the speaker is fully confident the states will carry out this election with integrity,” Ryan’s press secretary AshLee Strong said in an email to BuzzFeed News when asked about Trump’s claims.
Is he giving himself an out with that talk about "the states"? ("I hoped the states would conduct the election fairly, but Pennsylvania and Colorado look a little hinky...")

Are Ryan and McConnell going to back up Trump and his crazies if they cry fraud? I think they'll feel constrained if Clinton wins in a blowout (so get out and vote for Clinton, dammit). But it's a 4-point race now, and we're one Clinton coughing fit away from a 2-point race. If Clinton's Electoral College margin of victory is based on one or two close states, I don't trust the rest of the GOP not to endorse a coup.


AllieG said...

Steve, you are an excellent analyst in many ways, but your habit of cherry-picking polls that reflect your pessimistic view is a bad one. Averages, please.

rclz said...

It might be a 4pt lead nationally but that's not the way we elect our president.

As of this morning 538 electoral votes:

Clinton 322
Trump 214

That said, it only works if we get out and vote.

Ernest Miller said...

We should not take any comfort in Speaker Ryan's words. As usual, he is trying to have it both ways. He is merely trying to give himself plausible deniability for supporting Trump's paranoid ravings. Unless we hear clear denunciations of the claim that the system is rigged, Republicans are fully responsible for any resulting violence after the election. They need to speak out now because Nov. 9 is too late. Of course, the Republicans are unlikely to speak out against a meme they have been promoting for years.

Victor said...

Yes, in the last few years, elections have been rigged!

In 2000, they took thousands of people off the voter rolls in Florida, to help Jeb's brother, Dumbaya Bush.

And since then, they've passed fraudulent "Voter Fraud" laws in many states to suppress the votes of "undesirables" to the GOP:
Minorities, women, the young, the old, and non-"Christians."

And, I don't want to talk about electronic voting machines flipping votes to Dumbaya in 2000 and 2004.

Phil Freeman said...
This comment has been removed by the author.
bowtiejack said...

" . . . I hope we can start a coup . . ."

Didn't these jackasses try that in 1860? How'd that work out ?

Anonymous said...

Tried it 1934 as well. Fascists. NAZIs. Later one of the ringleaders, while financing Hitler, was "elected" to the Senate, and later still his son and grandson the presidency.

Characteristic number fourteen of the Fourteen Defining Characteristics of Fascism: Fraudulent Elections – Sometimes elections in fascist nations are a complete sham. Other times elections are manipulated by smear campaigns against or even assassination of opposition candidates, use of legislation to control voting numbers or political district boundaries, and manipulation of the media. Fascist nations also typically use their judiciaries to manipulate or control elections.

We have met the enemy,
Ten Bears

Philo Vaihinger said...

How do you do this in the US of A?

Is this coup supposed to happen between the election and the inauguration?

Who will carry it out?

Is the government, or anyway the military, supposed to at some point depose and arrest Obama, arrest Hillary, and invite Trump to the White House?

Have you been binge-watching Seven Days in May, again, like some kid playing his favorite song over and over and over?

Steve M. said...

Averages, please.

The RCP average for the 4-way race shows a Clinton lead of 5.5 right now. The Pollster average shows a 6-point Clinton lead. Big, but not massive, and smaller than Obama's '08 win.

I think Clinton has a huge GOTV advantage, and I think Trump will sabotage himself again. But in order to avoid fraud talk from people outside the Trump circle, I think it will be necessary for Clinton to have an Electoral College margin so big that you could remove her three or four closest states and she'd still win.

Steve M. said...

How do you do this in the US of A?

I keep wondering what would happen if congressional Republicans threatened not certify the results of the Electoral College vote -- the scenario Michael Moore addresses in the opening of Fahrenheit 9/11. If one close state puts Clinton over the top, I don't rule that out from McConnell and Ryan.

I also can imagine Trumpers identifying and threatening Clinton electors, especially in close states. There are a lot of guys with guns in Pennsylvania.

DerFarm said...

Trump couldn't effectively organize an orgy for Marines in whorehouse.

A coup attempt? Where HE might get hurt? Fuhgedaboutit.

Feud Turgidson said...

It won't be 4 pts and it won't be 12 either.

There's been no growth in Trump's support & there's zero basis for thinking that'll change.

Sam Wang's PEC analysis shows a Bayesian drift of 98% working in HRC's favor. That's a level Obama never reached this far in advance of Election Day, in 2008, leave aside 2012. That "stability" that's been in the data from the start, certainly from January, is now growing as a theme in the MSM. And that's the sort of story that feeds into itself, especially when the loser continues to do things and things keep coming up about himn that feed the perceived intention to show up to prevent his election. There's absolutely no signs of complacency in HRC's GOTV program.

Vote count theft likely has little to no application to some states (Alabama, Mississippi, Kansas, Wyoming etc), because it's already so deeply socialized as almost a fact of life. But in swing states, it's actually difficult to steal a close election AND get away with it. And each additional state vote count theft makes successful state execution, that is, without leaving evidence of it behind, EXPONENTIALLY more difficult. This is especially so where the D candidate is going to win, given the kind of aggressive election fraud investigations we've come to expect from the DoJ in D administrations.

The only way it's worthwhile attempting is in a really close contest at the national level, as in 2000 & 2004. There's just too many people involved in the process who have to be pulling in the same direction, & that's not what's been going on in the GOP.

Jim Snyder said...

@Feud: you might explain the significance of "Bayesian drift". Not everyone has taken grad courses in probstat / measure theory.

jsrtheta said...

It is exceedingly difficult to steal an election, period.

I have worked in an official capacity in many, many elections. People who talk about stealing votes at polling places are loons. Oh, I suppose if you are talking about an extremely small town where everyone pretty much votes the same way, you could clip a few more votes, but why bother? Otherwise, in any sort of populous area, and especially in a biggish city, it is impossible. There are just too many eyes watching, and representatives of both major parties are present for the whole ordeal. Add cops, prosecutors, federal poll watchers, volunteers and it becomes truly impossible.

Vote fraud via absentee voting would be easier, but even then you are talking a very tough row to hoe. The paper trails are long, and the opportunity to defraud is low.

When I was a prosecutor in Chicago, we worked every election. And at any time of day a precinct could have not only the election workers present, but election volunteers, party judges, Assistant State's Attorneys, Assistant Attorneys General, Assistant United States Attorneys, watchers from Project LEAP ("Legal Elections in All Precincts"), Chicago police, and representatives from any number of campaigns.

In the small towns of the Rockies, you have far fewer people monitoring things, but then you have far fewer voters, too.

The Trumpalos think the world is the way Alex Jones says it is. But then, they rarely bother to learn just how these systems work. And they largely work through the efforts of neighborhood people in the end. Trump himself hasn't a clue about any of this. One of the great videos I saw this cycle was of Trump inviting some cameras along to watch him vote a few years back. He didn't even know where his precinct was! IIRC, he went to three or four different polling places, none of which had him registered there, and he wound up casting a provisional ballot. He seemed a bit shocked that he couldn't just cast his ballot anywhere he wanted because of who he was.

Feud Turgidson said...

Jim Snyder - Think stability.

Of what? Of model being used.

To a certain extent, what you're asking of me to overcome isn't all that easy. For example, this paper is
considered fairly straight-forward Bayesian analysis.

OTOH, at some point in every one of the election cycles he's worked thru in real time (04, 08, 12, & this one), Wang posts a simplified explanation:

Note this passage:
'The Bayesian-win probability listed in the banner uses polls over the entire 2016 campaign to set a prior expectation for where things are likely to head. The second assumption also has the more traditional name of “regression to the mean.” '

Wang explains variously that the TIMING of the drift analyses he applies to polling data is critical in determining % likelihood of one candidate over another. Two implications:
1. Note that it's ONE candidate over ANOTHER one.
2. It's not possible to arrive at 98% adherence to the model on July 25, but IS possible (depending on what the aggregation results comE up as) the closer one gets to Election Day.

Today is October 16. Oct. 16 proved far too soon to arrive at this level of prediction in each of the 3 previous cycles where Wang (& others in the academic world) applied these methods. 2004 was more volatile right up close to end of polling. It proved too soon in 2008, by about 10 days. In 2012 the data didn't arrive at this current level of stability until about a week before the election.

Remember that in 2008 the conventions were held more than a month close to Election Day, McCain's choice for veep caused ripples in the polling data, and there was that little thing some called Collapse of the Financial System.

And in 2012 the model going into the fall debates was already supporting a closer reult than in 2908, and the debates themselves caused data ripples as Romney was widely perceived as having won the first, with Obama not regaining his pre-debates edge until the last one.

Again: the key concept is STABILITY of the MODEL, using early polling as the basis for establishing that model, and sticking with that model for however long the data adheres to it.

This cycle, there's just been hardly anything that's happened that rationally COULD disturb the model that in fact ended up doing much of anything. For example, the notorious 'post-convention bump' showed up for both nominees, but for Trump it was notably tiny and fragile, whereas for HRC it was high within the expected range proved at least partly sustainable (It appears to have helped keep her in the lead during her rough patch from Sept. 1 through the first debate).

Where I might have screwed up here in explaining the concept is no reflection on the adequacy of Wang's simplified explanation, or on the higher math. Indeed, one feature of reading the PEC comments is that the higher math heads chirp in regularly to debate whether Wang's latest analysis has succeeded in staying true to the model or, more often, whether the model's still holding it's stability. Sometimes I can keep up with those comment threads and sometimes I can't, & indeed occasionally Wang has taken up & adopted a criticism; but, so far at least, none of the objections on either front have won out over his methodology or the gross message the data is carrying.

Sorry for not responding earlier: Dodgers at Cubs tonight was stable by Kershavian Drift, yet too close to call until the last batter.

Jim Snyder said...

@Feud: Thanks! re "not easy to overcome", agreed, but well described (IMHO - I am not a probstatistician, so caveat emptor.)

Perhaps some intuition might be gained with a couple of sentences about Bayes Theorem?

It's implicit in your comment, but you don't explicitly describe the trade-off between duration of measurement and uncertainty which (I think) justifies the use of early polls to set initial estimates of "candidate performance" - you get a reading, but one accompanied by large uncertainty; the more readings that come in, the smaller the uncertainty ... with the caveat that changes in the underlying distribution have the potential for upsetting the statistical apple cart.

What I am still not getting is the implication / connotations of "drift"... not a term I have encountered in this context (except at Wang's site.)

Does "drift" mean "very gradual but real movement", or does it mean "random noise around the underlying real preference distribution"?

In casual conversation "drift" usually implies a slow but real change.

It seems to me that there are two - perhaps three - complications in these forecasts.

First, noise inherent in the limited samples used to estimate the underlying distribution, which noise derives both from statistics and from biases in the sampled population. (I can see calling this either one issue or two. Both would go away, of course, if you measured the entire distribution. Heck, you could identify a third noise source which wouldn't go away even with complete measurements: truthfulness of respondents, and the efficacy of poll questions at eliciting accurate indications of the underlying distribution.)

Second, temporal changes in the underlying distribution.

Guessing that "Bayesian drift" deals with temporal change of the underlying distribution ... but that's still not clear to me. Didn't find an intuitive explanation in the few minutes I spent looking on the 'toobz, but I'm sure one's out there.

Putting this out for your reax.

One other issue that isn't entirely clear to me is when & how you decide that the data no longer supports the model (notably, when the underlying distribution has changed.)

Frex HRC held her post-convention bump for something like a month, maybe six weeks. IIRC a lot of people felt after a month that the bump had "reset" the underlying distribution... until Trump gradually pulled even with her during (roughly) the following month.

One could argue that HRC's bump was real, and that events (e-mail & Foundation controversies) took her down ... or one could argue "regression to the mean."

Ditto Trump, one could argue that when Trump subsequently dropped like a rock following his 3 (?) weeks of self-created turbulence, the change in the polls reflected a real change in the underlying distribution ... or alternatively reflected a regression to the mean.

Or some of both, perhaps ...

But you're not writing a treatise on probstat. :-)

Thanks again for the explanation...

Feud Turgidson said...

Jim Snyder

1. The PEC approach uses "Random Drift" as a verifier for "Bayesian Drift". Wang is openly explicit about that. There've been questions raised in reader threads at PEC about whether that check is sufficient, but the questioners keep running into the problem that Wang and his fellow pure aggregation devotees have a track record and the doubters don't.

2. There's nothing preventing you from joining in on the reader comments at PEC. Wang has a pretty clear history of knocking out comments that are off-topic or come from trolls & spammers, but the things you're interested in don't fall within those categories. Having read thru a number of those threads, it seems to me likely that if you were to pose your questions there, you'd find yourself being answered by some of the regulars, not unlike at science oriented blogs. I myself have posted comments there that were shot down like a paper flivver in jet fighter dogfight, and I've had some comments knocked out as too off topic. But the wounds to my avatar didn't prove fatal.

Jim Snyder said...

@Feud - I'm not questioning the approach, even less am I questioning the results. I suspect my lack of comprehension has more to do with the terminology than math. Frex I have no idea what the sentence "Random Drift as a verifier for Bayesian Drift" means. If you (meaning "I") don't have the wordbook, you (meaning "I") will have difficulty extracting meaning from the sentences.

And yeah, I could post questions at PEC. Perhaps after the election ... I'm already spending way too much time watching Trump's antics. I should read up on probstat ... often felt that a deeper acquaintance with probstat would be useful in my day job.

Thanks again for your time. I dare say your post has been helpful to others as well.

Jim Snyder said...

@Jeff Ryan: useful post, thanks ...

Jim Snyder said...

[ROFL] Once in a while my day job requires that I read archival pubs. Feast on this, sheeples:

The article concludes that the rate of non-citizen voting in the United States is likely 0.

Not "approximately 0", not "within the MOE of 0". Nope ... *zero*.

Not that anyone should be surprised. Ahm jes' sayin'... I rarely see snark of this calibre in archival pubs.

Suck on this, Trumpkins.

h/t LGM

Glory said...

Hello, I'm here to introduce someone to you all, his name is Dr.Ekpen Temple a spell caster that help me restored my broken relationship, I saw an article on the Internet someone talking about him how he help her in her relationship, today I'm a beneficial of that article, so that is why I'm also talking about how he has helped me so that someone out there that is facing the same challenge can also contact him for help. Here is DR EKPEN TEMPLE contact info: ( or on Whatsapp number 2347050270218.