Why do political opinion polls seem like rubbish? PLUS the First Degree US election Special PLUS the latest release from DJ Dr Fish​

I have seen many electoral events where opinion polls uniformly over many months or years say the same thing. The Brexit polls, Boris Johnson’s landslide victory, Trump’s success in the US Presidential race in 2016 and Scott Morrison’s triumph in Australia in 2019 are recent examples of where opinion polls predict one thing yet on polling day a completely different result occurs.  Why is this?

The statistical theory is more damning than you think…

The problem is far more serious than most people understand.  There is very strong statistical sampling theory that says a relatively small sample from a large population will approximate the true population with, say, 95% confidence. Suppose that the preferred candidate has 53% support from the population of 200million voters.  Then a sample size of 1000 respondents should deliver an estimate of this mean between 50% and 56% and achieve this 19 times out of 20.  The probability of the sample mean falling outside this range is not a coin toss.  The probability of 2 polls being wrong is 1 chance in 400, not 1 chance in 4.

Therefore, if a succession of polls say the same thing but are wrong, the statistical likelihood of systematic error is negligible. Put anther way, polls are not worth the paper they are printed on.  

Sampling bias is a cop out…

The polling organisations blame sampling bias for their errors.  By this they mean that their respondent selection mechanism systematically omits a critical component of the population.  The favourite explanation for the US 2016 error is that dumb white people who voted for Trump don’t have phones.  I personally don’t buy this sampling bias argument due to the long term consistency of polling errors and the cross-section of consistency of all polling organisations saying the same thing. 100 polls with 1000 respondents each giving the same result is statistically equivalent to capturing the population in its entirety.  So how can the opinion polls be so wrong?

Maybe the opinion polls are right?…

My theory is that the polls are not wrong at all.  In fact, they are actually incredibly accurate at the time they are taken. When asked what their voting intention is a few months or even the week before an election, the respondents answer honestly.  My theory, however, is that the small number of swing voters don’t actually decide who they will support until they get into the voting booth. Once there, the gravity of the situation suddenly weighs heavily on their conscience and they do something different to what they told the phone pollster 5 months earlier or even the day before.

I support my theory with the fact that exit polls are incredibly accurate – “who did you vote for?” is a factual question as opposed to “who do you intend to vote for?”. The exit polls that were announced as soon as voting stopped over Brexit, Boris’s landslide, Scott Morrison’s underdog triumph and many other election events were shockingly correct. This suggests people go in to the voting booth thinking one thing but exit having done another.

Underdogs rule…

Related to this propensity to decide who to vote for while standing in the voting booth is the advantage of being the underdog.  If opinion polls uniformly predict one candidate as favourite then why not vote for the other guy? It can’t hurt can it? Achieving hot favourite status going into an election is a real handicap.  I vividly recall the Victorian Premier Jeff Kennett in Australia in 1995 entering an election with a massive 85% approval rating from the opinion polls.  How could he lose?  He lost simply because voters expected him to win!

First Degree’s US Election Special : Biden v Trump…

All this brings me to the Biden-Trump rumble that will be decided shortly. Biden and his supporters are in big trouble.  The political opinion polls are uniformly predicting Biden to be elected next Tuesday.  The pre-poll voter turnout is massive and from all reports the majority are in favour of Biden since they are registered Democrats.  Were I a ‘…thoughtful swinging voter…’, I would walk into the voting booth next week expecting Biden to win and probably expecting to support the Blue Tide.

But what has Trump actually done?  Unlike every other politician, he delivered on his promises.  He built a wall, he pushed back on China, he fought Congress – he turned out to be the maverick that people voted for 4 years ago.  Biden has “a plan” for everything but what has he done?  50 years in elected office and nothing – no bridges, no tunnels, no foreign policy successes, nothing.

What’s in it for me?  Trump wants to cut my taxes, protect my job and he talks my language.  Biden has a plan…am I going to pay for that, and why does the New York Times like him so much?

What will happen in the voting booth?…

The US election is all about Trump.  A vote for Biden is not excitement about his vision for America blah blah blah…it is a vote against Trump. The voters know Trump…and despite what they may have said before they arrived to vote, there’s no compelling reason to vote against him.

Trump will get re-elected. 

…and don’t forget the latest release from DJ Dr Fish

Do you like what you read? Then subscribe to our blog below…