Friday, October 14, 2016

The polls are worthless

Rick emails to point out an analysis of the "outlier" poll that has shown Trump consistently ahead of Clinton.  It uses a different methodology than the others, one that may control for the unique unknowns of this campaign better than the others.

Or not.  We'll see.  Personally, I think that all of the polls  are worthless, for two reasons:

1. We've known for months that there is a significant Bradley Effect in play in this election cycle.  When people think that their voting preferences will result in social consequences, they conceal their true choices from pollsters, throwing the polls off.  We see massive social consequences for Trump supporters in many areas - being labeled "racist" or "deplorable", having signs stolen from yards, etc.

We don't know - and cannot know - how big this Bradley Effect will be.  The Brexit campaign polls were off by almost 10% for exactly this reason.  If the effect is that large here, we're looking at a Trumpslide.  But the size of this is entirely unknown.

2. We don't know what voter turnout will be.  We see lots of enthusiasm among Trump supporters, with Trump rallies being measured in the tens of thousands.  We don't see that for Clinton.  On the other hand, Clinton likely has a well oiled Get Out The Vote effort, one that the Republican Party seems to be trying to deny to Trump.  How many Trump supporters turn out is unknown, and unknowable.

Both of these factors mean that the inputs to the polls (all the polls) are based on what's essentially conjecture.  Since the inputs are garbage, the outputs are garbage.  The numbers are worthless, because nobody knows the key factors that will determine the outcome.

But it's worse than this, actually.  The polls are worse than useless because almost all of them are being used to shape the public's perceptions of who is winning.  The pollsters - like the Media - are pushing one side; they've essentially weaponized the polls.

I plan to ignore all the polls because I simply don't trust the motivations of the pollsters, or the assumptions regarding respondent truthfulness or turnout.


Roy said...

Of course we all know that the only poll that counts is on November 8th.

However, the one thing that, as a Trump supporter, gives me pause, is that we were saying the same things about the polls back in 2012. It turned out that they were mostly right.

Borepatch said...

Roy, true about 2012. But this election is nothing like anything I remember. The rules have just seemed not to apply.

SiGraybeard said...

The cynic says that the polls now are being used to make opinion not show opinions. Pollsters and the news business go hand in hand, and the media's corruption for Hildebeest and the Evil Party is beyond dispute, thanks to Wikileaks.

On the other hand, firms like Rasmussen and the other professionals sell their services based on accuracy. In the closing moments before the election, they'll try to get it right so that they can advertise they got it right. Those will be the only polls worth looking at.

Guffaw in AZ said...

Yeah, what Roy said.

Assuming there's no voter fraud...


Mike said...

The other item that makes me question the relevance of the poll data is the near-certainty of widespread voter fraud in this election. I only see that going one way, and even if Trump does manage to capture the true popular vote that might well be enough to offset it. We can also guarantee that there will be no investigation or prosecution of fraud.

Couple that with the fact that informal conversations have shown me that the video of Trump's comments from a few years back plus the new accusations of groping, etc. have actually turned off a number of people who were going to vote for him, and the polls may not be as far off as it seems. Most of the people whose minds have been changed have simply said they can't vote for either candidate in good conscience and intend to stay home, which in practice will be a vote for Hillary.

Archer said...

@BP, regarding point #1: It's true we don't know the magnitude of the Bradley Effect (a.k.a. the Wilder Effect). We also don't know the direction; the polls leading into the election don't always favor the politically-correct candidate (read at that InfoGalactic link about the 1989 Virginia gubernatorial election -- the "Wilder Effect" namesake -- where the polls showed landslide-level support (greater than 9% lead) for the white candidate; in the end he still won, but only by less than one-half-percent.

Given that, the Bradley Effect could go the other way, and this election could be a Hill-slide.

@Mike: It will be interesting to see what happens if Hillary wins, but by less than the "margin of fraud". What if she wins by 0.1%, but analysis shows voter fraud on the scale of 0.2% (with it already well-established that statistically, dead voters overwhelmingly favor Democrats)? What will the country's reaction be?

It won't be good by any measure, but it will be interesting....

Jeffrey Smith said...

I think the Bradley effect can operate in favor of Clinton. People have read about all those alt right folks sending out death threats, and are ashamed to admit they intend to vote for one of the most corrupt politicians in American history.

But overall I agree with you. In fact I think polls are usually junk even when they are more accurate. Sample size too small and demographic assumptions to balance that out are too speculative.