As the 2016 election draws nearer, polls continue to fluctuate wildly. Some have Clinton firmly in the lead, others have put Trump narrowly ahead.
But it’s not the first time polls have been contradictory, and come Wednesday, you shouldn’t be too surprised if they prove wildly off the mark.
In Obama’s 2012 re-election campaign, some polls had been consistently predicting a loss for Obama.
Right up till Election Day, most pollsters had Obama and Romney neck-and-neck.
The final result?
Obama won by 5 million votes. The most decisive win for an incumbent president in almost 50 years.
One of the main reasons American elections are so difficult to predict is voluntary voting.
As well as estimating who people are going to vote for, you also have to guess whether people are going to actually going to show up and vote.
In Australia if you’ve got a big enough, representative enough sample size, you can be pretty confident of your accuracy.
In America, there’s a bit of guesswork given that only about 60% of registered voters will actually show up.
That explains how polls taken at the same time can be wildly contradictory, and why campaign-watchers will rely on averages between polls rather than any single piece of data.
But even averages can be off.
In 2008 Obama’s landslide win was driven by high turnout in minority groups, such as African Americans, who traditionally recorded poor turnout.
Barack Obama and his wife, Michelle, left, and Vice President-elect Joe Biden and his wife, Jill, celebrate after Obama's acceptance speech in 2008 (AAP). Source: AAP
In 2012 pollsters were uncertain whether those same voters would turn up again.
There was no historic ‘hope and change’ campaign to energise them, the conventional wisdom said.
But show up they did, thanks in large part to a sophisticated, data-driven ‘get out the vote’ effort from the Obama campaign.
Hundreds of thousands of Obama volunteers made phone calls, knocked on doors and even drove voters to the polls on Election Day.
It’s those efforts which can distort the electorate and throw up surprises.
The question for pollsters and campaign-watchers this year will be whether Hillary Clinton is able to energise target demographics in the same way as Obama.
Women, Hispanic voters and African Americans favour Clinton – men, particularly white men, favour Trump.
For what it’s worth, my money’s on a highly successful ground effort for Clinton.
The sophistication of the Democrat’s well oiled ‘get out the vote’ machine – built over two successful Obama campaigns – now goes well beyond the appeal of any particular candidate.
If Clinton wins, it will be minorities and women – and the Clinton campaign knows it.
It’s why the rhetoric on Trump’s record of insults and assaults on women has been so fierce.
It’s why in the swing state of Florida Democrats are that their voter registration drives have swung the electorate from 67% white to 64% white.
But there are other issues which complicate US election forecasts as well, such as the lack of quality polling conducted in crucial swing states.
Not knowing exactly who is going to show up on election day, or who is going to vote by mail, makes it difficult for pollsters. Source: REUTERS/Aaron Josefczyk
With the Electoral College system, presidential candidates compete to win states which have a set number of votes based on their size – it’s a winner takes all system in each state.
That means the entire election can come down to a few swing states – meaning that while nation-wide polling will show trends, it’s much less useful than dedicated polls in swing states.
The problem is, nation-wide news outlets based in the media centres of New York and California don’t appear to be that interested in spending the resources on polls dedicated to single states.
This year, the most important swing states are Florida, Pennsylvania, North Carolina, Michigan and Virginia – followed by Colorado, Wisconsin, Ohio and Nevada.
While North Carolina and Florida have seen frequent polls – which reveal both to be toss-ups – there are hardly any recent, quality polls in states like Michigan, Pennsylvania, Wisconsin and Colorado.
Right now, the campaigns will be running targeted polls themselves - and keeping track of millions of people who have voted by mail - but they’ll only publicise those numbers if it’s to their advantage to do so.
Because they know who they're targetting and who is voting early, the truth is that the campaign managers themselves probably have a far better idea of how Tuesday will turn out than you, me, or any of the public pollsters.
