US Presidential Election – How Reliable are the Polls?

Is Clinton’s Lead Over Trump as Large as Advertised?

Once upon a time, political polls tended to be pretty accurate (there were occasional exceptions to this rule, but they were few and far between). Recently there have been a few notable misses though. One that comes to mind is the Brexit referendum. Shortly before the vote, polls indicated the outcome would be a very close one, while betting markets were indicating a solid win of the “remain” vote. The actual result was around 52:48 in favor of  “leave”, so this was quite a big miss.

Another case that confounded even the most seasoned forecasters was the race for the Republican nomination. See for instance this article by Mish from January, in which he rightly berates famous election forecaster Nate Silver for vastly underestimating Trump’s chances to win. Silver held them to be around 12% to 13% at the time, which turned out to be a miss of truly monumental proportions. He kept missing the mark for many more months to come (essentially until the point in time when Trump had made the transition to “inevitable nominee”).

We currently follow press coverage on the presidential election only cursorily, for lack of time, and also because it seems both very superficial and one-sided. The mainstream media bias in this election seems astonishingly blatant. To quote Stefan Molyneux, there is apparently no mud one cannot sling at the orange-haired maniac. We are of course well aware that Donald Trump has an uncanny ability to put his foot into his mouth, but it seems almost as if little else is discussed (skim through a few MSM articles at random and see how much you learn about his policy proposals – good luck).

NYT National Polling Average Asserts

On the other hand, it is not as if Hillary Clinton were widely considered an angel. The woman has spent decades mired in scandals, from her credulity-stretching career as a cattle futures trader in the 70s, to the Whitewater affair, Travelgate, the harassment of women pointing fingers at her philandering husband, to the recent questions about the Clinton foundation, the gigantic speaking fees she gets for mouthing platitudes at corporate gatherings for 20 minutes a stretch, the  e-mail server controversy (and not to forget, her sociopathic streak)… and we’re sure we have left a few things out.

In short, she is just as polarizing a figure as Trump is widely held to be. Then there is the fact that Donald Trump continues to draw huge crowds wherever he goes to speak, a feat Hillary Clinton is not known for.  Perhaps this is merely a sign that Trump voters are more energized for their candidate – but that seems quite an important factoid by itself. Remember the Brexit post mortem in this context: the “remain” camp mainly failed because it was unable to create even the slightest bit of excitement for its cause.

All of this made us wonder why Hillary Clinton is so solidly ahead in political polls. Below is a chart of the NYT’s “national polling average”, which concludes that her current lead boils down to an “89% chance of winning the presidency”. Really? That sounds a bit like Nate Silver judging Trump’s chances to win the Republican nomination back in January, i.e., it is a bit hard to believe.

If one looks at the list of polls the NYT uses to construct this average, one can see that Clinton’s lead actually varies quite widely among them. Some of the organizations and/or media conducting the polls either have a known political bias, or a known bias in this particular election. The reported size of her lead seems to be correlating rather closely with these known biases. This is slightly mystifying, is it not?

Voter Excitement

Most of these polls are conducted via telephone, but there is also an online tracking poll by UPI/CVoter.  Its margin of error cannot be calculated because the  participants are self-selecting, but UPI mentions a credibility interval of 3 percentage points (apparently based on the Bayesian approach to quantifying uncertainty). This particular poll has fluctuated in quite a wide range, but at the end of last week, Trump has overtaken Clinton again for the first time since July:

Once again, it is possible that Trump voters are simply more engaged and more likely to self-select for this poll. Trump certainly packs more internet punch than Clinton through his Twitter feed, Facebook and You-tube presence (even though his tweets occasionally land him in hot water). As noted above, Trump has something in common with the erstwhile Democratic rival of Clinton, Bernie Sanders: he is generating genuine excitement among his supporters.

We wonder what it takes for Democrats to get excited over the prospect of voting for a woman who bears a large share of responsibility for turning an entire country into a war-torn failed state overrun by terrorists and is showered with money by Wall Street (note: she not only gets serious money for her probably sleep-inducing speeches from the likes of Goldman Sachs, she has actually received 2,550 times more money from hedge funds than Trump).

Presumably many are clinging very strongly to the “lesser evil” fantasy, but surely they realize that she stands for business as usual (including more war) – the quintessential representative of the cronyism characterizing Washington.

What Might be Skewing the Polls?

One thing that occurred to us already during the nomination process was that certain Trump supporters were likely to refrain from identifying themselves as such, given the steady drumbeat of condemnation of their man in the media.

What we didn’t know was that the phenomenon actually has a name, which a friend recently pointed out to us. It is called the “Bradley effect” (you can read all about it here), named after the supposedly winning (but actually losing) candidate in the 1982 gubernatorial election in California. Even the exit polls turned out to be wrong in this case.

A short while ago we have come across another possible explanation. Below is a video in which Stefan Molyneux speaks with Bill Mitchell. It has to be pointed out that Mitchell is a Trump supporter, so his views are presumably colored by his own bias to some extent. Nevertheless, he presents very interesting arguments.

Apart from mentioning some of the things we discuss above (e.g. the enthusiasm Trump generates), and the likelihood that independents and currently still undecided voters will overwhelmingly vote for Trump (which incidentally would also be in line with the “Bradley effect”), he notes that there are certain technical problems with the polls.

Specifically, he points out that the polls are skewed because Democratic voters are generally oversampled and the results are not corrected for that. He mentions a few counter-examples that seem to confirm this (note that we have not fact-checked these assertions).

While the term “social mood” isn’t mentioned in the course of the conversation, Mitchell does point to it when he says that (we are paraphrasing) “Trump is just riding the wave of an already existing trend”. This trend is noticeable all over the world as we have previously discussed (see e.g. “Incumbents Swept from Office Around the World”). Here is the video – it is well worth watching:

Stefan Molyneux speaks to Bill Mitchell about the polls and Trump’s chances

Trump to Black Voters

Part of the “Bradley effect” would also consist of demographics that are thought to be overwhelmingly favoring Clinton containing more Trump supporters than is generally thought. And there may well be many who are prepared to reassess their views when Trump tries to reach out to them – which he appears to be doing of late, once again employing a theme tied to the current social mood.

Consider this brief video clip from a speech he recently delivered at a rally in Michigan, in which he explains why African-Americans should consider voting for him. His argument is basically: you have it bad now, so what do you have to lose? Better vote for change (in the form of Trump), instead of voting for the status quo again.

This idea may have unexpected appeal, considering that blacks have overwhelmingly voted for Obama and have lost a great deal of ground economically under his administration. During his presidency, black home ownership has declined from 46.1% to 41.9%, while black food stamp recipients have soared from 7,393,000 to 11,699,000 people. One statistic that was improving – namely unemployment, and specifically youth unemployment – has come undone again after the introduction of higher minimum wages, rising from 28.1% to 40.1% in a single month.

Trump asks African-Americans to consider voting for him