US Election: The media misreport and then blame the polls - it's an inexact science
Part of this inexact science means there is also plain luck involved in any election or referendum
For an election junkie like me coming off a US election night coverage marathon is like coming off a drug. The drama was exhilarating. The analysis, forecasts , projections - great. Frankly I think we should have an election like that and Brexit every six months or so - just for the fun of it.
As usual of course the mainstream US and UK media had been subtly pushing a pro-Clinton line, and then when it goes wrong they dump on the polls. If Clinton had won - the media, including the BBC, would probably have been trumpeting their pundits and the polls wouldn't have had a mention.
It's amazing the power of the main media e.g. the BBC - they can broadcast distortions about the EU Referendum polls and it just gets constantly fed into people's psyche, so they end up believing it i.e. the EU Referendum polls said the UK will vote Remain, and were wrong. The UK EU Referendum polls never said the UK would vote Remain.
They basically said the result was too-close-to-call with the online polls tending slightly towards Leave. Even here in N. Ireland LucidTalk published their last pre EU Referendum poll 3 days before the referendum (Belfast Telegraph and The Sun) forecasting basically what turned out to be the final result at 56% Remain to 44% Leave - of course this wasn't published by BBC NI probably because it was too pessimistic about Remain.
They're at it again with the US election - if you listened carefully it was the media and pundits that predicted a Clinton victory not the polls, - the polls always said Trump had a good chance. Plus it was the pundits and TV networks which did all the US Electoral College projections (the US' unique system that elects their president), not the polling companies. Projections of the Electoral College are precarious and have large errors (though they're great fun - I know).
Cutting to the chase - The US national polls predicted a narrow Clinton victory i.e. in the total national vote and this turned out to be correct, as she's going to win by about 1 million votes in the end. Then when we go to the state polls - out of 50 polls, 46 were correct including crucially Ohio and Florida, two key swing states. Ohio had gone 3-4 points (i.e. comfortable) for Trump in all the polls in the last 2 weeks of the campaign, and Florida had gone for Trump 0-1 points in the last week. This last point is hugely important and should have been reported by the media - was it?, of course not, because it wasn't pro-Clinton.
A proper professional correct media report over the weekend before election day would have said 'Crucially it should be noted that two swing states are now trending Trump in the polls... this now makes it sort of 50/50 in the Electoral College, and though Clinton is still a slight favourite the latest polls in these two key swing are a big warning sign to Clinton..'. - That would have been a fair professional report. Instead we had the BBC correspondent in Washington saying (weekend before the election) that 'It's now accepted that Clinton will almost certainly win..' Ehh? Where did that come from? Even on the election results night we had Jeremy Vine doing his graphics (in front of a graphic of the Washington monument) saying that '.. the first shock was Ohio as against all the pundits and pollsters it went for Trump..' - that is just plain wrong. All the polls (and I mean ALL) had Ohio comfortably for Trump from two weeks out. At the very least that remark represented bad research.
Look, the total US polling wasn't great but it was OK/Passable, and again the warning signs were there e.g. Ohio and Florida. If you average all the polls, 4 states (out of 50) were wrong: Pennsylvania, Michigan, Wisconsin and Nevada. The first three of these are part of what is known as the 'Rust Belt' with the old, now mostly defunct, industries of coal, steel, and heavy manufacturing. Yes, the polls didn't get enough good representative sampling in these states (even though the very last polls did suggest Michigan was in play) and couldn't get enough access to that white, working class, angry, sort of left behind, demographic.
Just to show you the warning signs were there - here are four tweets from the LucidTalk Twitter timeline running up to the election (all date stamped)
— LucidTalk (@LucidTalk) November 6, 2016
As we said 2 days ago - one state could swing it & it could be Florida or Pennsylvania: https://t.co/d8DBNvEHkN
I thought last night that Trump had only one long-shot Electoral route to the White House - on tonight's analysis he now has 3-4 routes.— LucidTalk(@LucidTalk) November 3, 2016
Do you see the pattern?
I was invited onto BBC Northern Ireland's Good Morning Ulster after the US election - NB you notice not before, that's for all the in-bubble commentators that waffle on and leave so much wriggle room they could swim the Atlantic. The first question put to me was that (a) The 2015 general election polls were wrong, (b) The UK EU Referendum polls were wrong, (c) The US election polls were wrong. The EU Referendum polls were accurate (see above), and the US polling wasn't wrong - In fact, look at the post on LucidTalk's Facebook: 27th October, 21.15, as it forecast a Trump win based on the three US Poll companies we track (we also follow their methodologies). Yes, (a) was a bit out, but again if the media at that time had looked behind the data the warning signs were there e.g. Miliband's ratings compared to Camerons etc.
It's hard to get across to people that polling is an inexact science (i.e. they're a mixture of assessment, human behaviour and science). Part of this inexact science means there is also plain luck involved in any election or referendum and Trump was very, very lucky with the way the votes fell for him in the Electoral College system. Was there some higher power at work? - because if Trump and his team had designed the ideal structure for where they would like their national vote to be split in the key states, they couldn't have designed it better than what happened.
Just to finish (this rant) the BBC's William Crawley kindly sent me a link to a major respected US review of the US polling and maybe the last three paragraphs sum it up:
Pro-Trump operatives argued that even when some polls hinted at Trump’s strength, it was ignored or explained away by the media and analysts.
“Most of the press and folks in DC were science deniers when it came to this election,” said veteran GOP operative Curt Anderson, an adviser to a pro-Trump super PAC. “Even in the face of polls that showed it very close, they all said that Trump had almost no chance. It was because they couldn’t imagine it happening.”
He added that “they are in a bubble, and that bubble has just been burst.”
And finally, and to all of you out there knocking the polls including the BBC - you can't win. Regardless of the result we have far too many pie-charts, colour graphics, percentages and figures to prove we got it right - and that's even when we're wrong.
Bill White is Managing Director of Belfast polling and market research company LucidTalk.