The FT reports, in a fulsome and fairly scathing piece on the polling industry this fact:

There have been more than 900 polls by 141 pollsters on the US election next week.

If there is now that much polling, how come the pollsters still get it wrong?

It’s a good question that requires an answer, not least so customers including the PR trade understand what they are getting when they pay for a poll.

My conflict of interest here is blatant*, but look past it for a minute.

A few other takeaways from The Problem with Polling:

1) Hacks are terrible at numbers; they don’t understand polls, and reach for a simple conclusion even when there isn’t one.

2) Betting markets seem as accurate/better than polls. So why bother with polls?

3) Pollsters have (finally) turned away from phone polls in favour of online surveys, having noticed that only the slightly strange still have landlines and talk to strangers on them about their political opinions.

The case for the defence:

The polling industry is getting better all the time. It hasn’t quite got around the “career responder” issue – people who do nothing but answer polls all day are an unusual, unrepresentative bunch – but it’s working on it.

So many polls now are bang on, even on quite high levels of detail.

The ones we get wrong are the ones non-pollsters look at – what percent of the vote did the main parties get at a general election?

All major pollsters suffer from under-estimating the right-wing vote. Even when we adjust for that, we still get it wrong. This is a problem, depending on what you expect.

Polls are just snapshots in time, but people relying on them/paying for them, don’t get that, hence their frustration.

I think a distinction needs to be made between measuring what people think (opinion polling), and predicting their behaviour (voting intention polling). 

The latter is much more difficult than the former.

Whether someone acts on their opinion is a complicated science. The reason the polls were wrong last July was largely down to a failure to predict lower turnout. 

A load of people who preferred Labour just couldn’t be bothered to vote. Reform voters were the opposite.

Betting markets are definitely not perfect either, since they are quite easily distorted by, say, Trump supporters pumping tons of cash on a Trump win.

For others who commission polls, I promise it’s all smarter and better value for money than it was, unless you’re expecting us to get the exact Trump/Harris percentage vote next Tuesday.

In any case, the US election polls stand at 50:50 and winner takes all. Polls aren’t able to add much value and inevitable error (whether within the statistical margins or not) makes a huge difference.

For almost everything outside politics – people’s attitudes to Covid, to brands of shampoo, to electric cars, we’re getting really good at this.

So, 50% of people might love your client, and the other 50% loathe them.

What polling can do is reveal who those people are. Something that Advertising or PR frequently fails to do.

Like it or not, private polling done for Labour, and other political winners, use these polls, which never get made public, to work out what to say, and to whom to say it, to get the outcome they want.

It worked.

For those who are still sceptics:  Has an inaccurate weather forecast stopped you looking at the weather report on your phone in the morning?

*Simon is a partner at digital pollsters findoutnow.co.uk.

Please send candidates for press release of the day to: Simon.english@roxhillmedia.com