If I thought there was no magic number before...

Back in 2012, just about six weeks out from President Obama's re-election, I called the exact margin of the US popular vote to a decimal place. Without a spread sheet, without a fancy algorithm. I was actually just sitting on a train, having a chat with the random person across the table from me. The polling numbers were stable, predictable. I'd been working with a "poll of polls" for Research Magazine and writing weekly about what the numbers were saying. I was confident. And, as it showed six weeks later, I was right.

Now? I wouldn't bet a nickel on this election's outcome. Because I haven't a clue what is going to happen.

In the UK, we talk a lot about the "shy Tory" or "shy Leave" or "shy whatever" voter. I have quite a few theories on why our public polling is so unreliable. But that's for another post. Today's question is why is it that I don't trust the current US public polling, when it's been pretty good for the last few elections?

Simple: is this election like any other we've lived through in America? Have we ever had two candidates so disliked? Two celebrities running against each other? One spewing hate every which way and the other one hated herself by a sizeable chunk of the country's population for more than two decades? No. No we haven't. 

Why does this matter? As far as I am aware, most pollsters in the States are asking the same old questions they always ask, the way they've always asked them. Then, they are building their models based on turnout, likelihood to vote, favourability, demographics and a thousand other things, all of which are based on what has happened before. But this election is unlike one we've ever seen before.

The current polls are showing a popular vote win for Hillary Clinton, by margins of anywhere between 3 and 14 points. The electoral college math maps look like a sea of blue. In some swing states, we are seeing minorities coming it at 0% for Donald Trump. Our best soothsayer, Nate Silver, is now predicting how big of a landslide it will be rather than who will win. So yes, I get that we are on track for our first female President. But I can't get comfortable with it. I have a sneaking feeling that we aren't seeing everything in the numbers. That there is no way that no African Americans in Ohio are going to vote for Trump - his economic message must resonate with at least some minorities in the rust belt. But I suspect that they don't want to admit it in a poll. And if that's the case, how does that logic apply to other subgroups that the pollsters are analysing?

We don't have a history of shy voters in America. But Trump's message must resonate with more people than are comfortable owning up to it in a poll. And so, for the first time in my adulthood, I cannot for the life of me predict what is going to happen on election day in the US. Not the winner, not the margin, not what happens next.

Can you? Can you, really?


[originally posted on LinkedIn]

Why marketers may be getting consideration sets completely wrong...

No Magic Number