#Open journalism No news is bad news

Your contributions will help us continue to deliver the stories that are important to you

Support The Journal
Dublin: 2°C Saturday 28 November 2020
Advertisement

Don't be fooled: The polls weren't wrong in 2016, the analysis was

The goalposts – and polling analysis – have shifted since 2016, Noel Rock writes.

Noel Rock

LET’S START AT the top and save you some scrolling. It’s a Biden win next Tuesday, no question. We can see it in the polling, we can see it in the funding patterns, we can see it in terms of where campaign attention is being deployed and we can see it in terms of early voting. 

In 2006 and 2007/8, during my time working for Hillary Clinton, I developed a sense of what a winning campaign looked like in a U.S. context and how – often – much of what is portrayed to the media about what’s happening in a campaign isn’t really the reality at all. Sometimes the pageantry, fanfare, anecdotes and myths can be put to the sword by cold, hard data-driven analysis. 

It was true in 2008 as – while Hillary Clinton received enough votes to win a primary in any other cycle – Barack Obama had grown the electorate sufficiently such that the Clinton campaign were aiming for the wrong goalpost.

The Clinton campaign aimed for 16 million votes and got 17.49 million. Enough to win, right? No. They were up against somebody who had got 40,000 more votes, and in all the right places. Game over.

The same pattern is emerging here in the 2020 election. It is very probable that Donald Trump gets around the same number of votes as in 2016, yet comes nowhere near winning. We can see this reflected in early-voting turnout especially.

As the days tick down to the conclusion of the campaign, much talk has shifted to the truly eye-catching level of early voting.

This remarkable level of early voting is symbolic of two things: 

  1. A high level of voter enthusiasm on both sides – the latest analysis suggests in excess of 160 million votes being cast, a substantially higher number than the 122-138 million range of the most recent 4 Presidential elections.
  2. Covid-19 driving demand for postal voting

This would mean that, should voting hit a turnout of, say, 150 million, you could see Donald Trump get the same number of votes as 2016, maybe even exceed it, yet come nowhere close to a victory. The goalposts have moved.

As of Wednesday evening, over 75 million early votes had already been cast. Of these, over 15 million are new voters who had not voted in 2016.

Incredibly, three counties in Texas had – as of Monday – registered more early votes in 2020 than there were votes in total in 2016. One full week out from the election, that is simply incredible. By Wednesday evening, nine counties had.

In the state as a whole, Texas has now registered 8.2 million early votes – the equivalent of 92% of the total vote registered in 2016. It is probable that the number of people who vote in Texas in 2020 will surpass the number that voted in 2016 – before Election Day takes place. A truly remarkable step change.

While Texas is a leader in this regard, we see other states where the early vote registered is a high percentage of the total vote in 2016. Georgia, Arizona, and North Carolina are above the 75% mark now, while Montana has seen 81% of its voters vote almost a full week before Election Day.

But, what does it all mean? What is the net output?

For this, we should look to the polls themselves.

While some punditry still lazily relies on the idea that the polling was “wrong” in 2016, that simply is not the case.

What is true is that the punditry around the polling was wrong. Most polls predicted a Clinton victory of between 2.5-4% and, on polling day, a 2.1% popular vote ‘victory’ was registered. 

The problem? Clinton’s margin was in precisely all the wrong places. Donald Trump pulled off the election equivalent of ‘Moneyball’ and won by the narrowest of margins in all the right places – a 0.23% win in Michigan, a 0.72% win in Pennsylvania and a 0.77% win in Wisconsin.

A cumulative margin of fewer than 77,000 votes between the three states resulted in a crucial 46 electoral college votes falling into the red column rather than the blue column. That’s where his victory lived. That’s where her Presidential bid ended.

Meanwhile, Clinton ran up large victories in places like California and New York to no avail. The polling was right, but the margin was in all the wrong places, and the analysis, punditry and false confidence in the run-in reflected that.

So, a lesson of sorts has been learned. 

Punditry and analysis of polling has become far more detailed and fine-grained, and indeed State-level polls themselves are being commissioned for public consumption on a more regular basis (campaigns often commission many of these and hold onto them), along with District-level analysis so we can delve specifically into particular demographics. These are all good things, of course. 

Polling analysis is a bit like the Theodore Roosevelt adage – “a vote is like a rifle: its usefulness depends upon the user”. I believe 2016 has improved both the depth of polling as well as the understanding of the need for proper analysis. 

All told, though, we have a far more complete picture, and the picture is this: Biden is winning. I won’t take you through all the detail – but let’s duck under the hood for a second and see one factor which is powering this: over-65s. In 2016, exit polling data showed 52% of voters over 65 years of age going for Trump. This time around, the polling figure is varying between 38-41%. 

It is a substantial dip, and isn’t a flash in the pan: it has been this way since the start of the summer.

The handling of Covid will prove to be a decisive factor in this change. Approximately 10,000 Americans per day join the swelling ranks of the over-65 demographic, so this change for Trump, and the margin of Biden’s lead among this cohort, has an even more outsized role in an election than usual. This is, in short, a major problem for the incumbent.

In many ways, it has been a remarkably steady race – even if it hasn’t felt that way. The average polling lead which Biden has over Trump nationally has been incredibly steady – and wide – since July. In key swing states, the margins have grown and then contracted a little, but again the broad story is much the same. 

#Open journalism No news is bad news Support The Journal

Your contributions will help us continue to deliver the stories that are important to you

Support us now

In terms of states to keep an eye on, while there is a temptation to get into talking about Georgia and Texas, it’s more realistic to look at Arizona on election night as a new frontier for the Democrats – which would go Democratic for the first time since 1996. Prior to Clinton’s win there in 1996, you have to go back to 1948 to see that state turn blue. 

Similarly, if you are worried that the polls have got it wrong again, the early voting is illusory and Trump could make lightning strike twice, then it makes sense to cast your eyes on the states which played as the backdrop to Trump’s victory in 2016: Michigan, Pennsylvania and Wisconsin. Without these, Trump can’t manage it. The polling has Biden up strongly in all three. I wouldn’t bet against him.

Noel Rock is a former TD and previously worked for the Hillary Clinton campaign. 

voices

About the author:

Read next:

COMMENTS (86)

This is YOUR comments community. Stay civil, stay constructive, stay on topic. Please familiarise yourself with our comments policy here before taking part.
write a comment

    Leave a commentcancel