Thursday, November 10, 2016

Why the Pollsters and the Betting Markets Got the Election All Wrong

By Robert Wenzel

If any group should understand why the pollsters got the presidential election  wrong it is economists of the Austrian School.

An important tenet of the Austrian school is that the proper methodology for the social sciences is different from that of the physical sciences.

In the physical sciences, we deal with cause and effect relationships that can be isolated and do not change. For example, all other things being equal, if we poor x gallons of fuel into a rocket ship, we can predict exactly how far it will travel every time.

In the social sciences, we do not deal with such constants. There are no possibilities in the real world of human exchanges with "all other things being equal." The social sciences deal with human action where people can change their minds and valuations at any time, and act differently from one period to the next. Plus we are dealing with new and different complex inputs all the time on individuals. Given these circumstances, it is not possible to make the exact types of forecasts that are done in the physical sciences. Specifically, it is nearly impossible to make exact forecasts in the world of economics and occasionally difficult for election pollsters.

Here is why the pollsters got it wrong on Tuesday: Key variables changed from the 2012 election. It caused their models, which sort of operate as though they are in the physical world and constants exist, to get smacked up against reality.

The models probably had a correction factor for how many voters lie about who they are going to vote for. This was based on previous elections. But for this election, the models likely underestimated the number of people who said they were going to vote for Hillary but actually voted for Trump.

That was problem 1.

But here is the real kicker.

The pollsters ask registered voters if they plan to vote and who they plan to vote for.

They then, based again on previous election results, adjust their models for how many people they expect will actually show up at the voting booths.

They got this totally wrong for the Tuesday election. As I have pointed out here at EPJ and Target Liberty, Hillary Clinton couldn't get a wave going in a major league baseball stadium. She is just not an inspiring person that people want to follow with excessive enthusiasm (except for extreme nutjob lefties). Although her supporters told pollsters that they planned to vote for Hillary they just didn't turn out in the same ratios as occurred in 2008 and 2012 for Obama. This threw the models off. The models overestimated the number who said they would vote for Hillary and would actually turnout.

It was a different story with Donald Trump. His supporters are rabid. They likely showed up much more heavily based upon what they told pollsters relative to the turnouts for John McCain in 2008 and Mitt Romney in 2012. That is, if a Trump supporter said he was going to vote for Trump that voter showed up at the voting booth at much higher rates than the rates that occurred for  McCain and Romney. Thus, the pollsters underestimated Trump's turnout.

This is what blew up the models. They didn't properly measure the intensity of those planning to vote for Trump versus those planning to vote for Clinton.

As for the betting markets, they were a total disaster in making a correct forecast. They really don't do anything but follow popular opinion.

As Pippa Malmgren pointed out as voting was wrapping up on Tuesday:


I have written about the problems with those who view betting markets as some sort of magical predictor before.

Joe Osmundson was kind enough to put out this email after the Trump victory:
Care to give credit to Wenzel and his striking call on the Bettors market failure's yet?  Or are you just going to ignore good libertarian and Austrian economics analysis?

Does a good job here in making the case that a Bettors market can't be trusted 100% or any Poll:
There is an  idea in some economics circles, which sadly include some Austrian school economists, the idea being that bettors somehow give a sound indication on how an event will develop, with an implication that bettors somehow have better knowledge than the general public. I have long objected to this view. 
In my view, all that betting markets do is reflect the views of the bettors. In some cases, the bettors can be well informed, in other cases they may not have a clue.
Read the rest here.

 Robert Wenzel is Editor & Publisher of  EconomicPolicyJournal.com and Target Liberty. He also writes EPJ Daily Alert and is author of The Fed Flunks: My Speech at the New York Federal Reserve Bank. Follow him on twitter:@wenzeleconomics and on LinkedIn.

6 comments:

  1. I have always had suspicion of polls. Who are these people that take polls? I don't take surveys on a good day, because theyre a nuisance. I think a lot of people are that way. But who is it that has the desire or time to answer robocalls? Probably not a working stiff who has to work overtime to pay for the rising costs of his Obama care.

    ReplyDelete
  2. This is the first thing I thought of too. I conservative working stiff doesn't want to talk to the guy on the phone, especially with Trump talking "rigged system" all day. Progressives love that crap.

    ReplyDelete
  3. ─ Hillary Clinton couldn't get a wave going in a major league baseball stadium. ─

    She did get the majority popular vote, Robert, unlike Brexit.

    ReplyDelete
  4. It would seem that betting markets are so small and easily manipulated that they are themselves a form of push-polling.

    ReplyDelete
    Replies
    1. Re: Timothy,

      That is exactly what I was thinking. Ever since the US Senate and the government bullied Intrade towards not accepting US traders, the market became too restricted to be useful. This was noted by John Stossel (the libertarian journalist) who liked Intrade but who conceded that one the government imposed itself on these US-based prediction markets, the whole market became less reliable.

      Delete
  5. How about the accusations that the pollsters were over sampling Dems and other tricks to skew the results in the favor of Hitlery.

    ReplyDelete