[leonk]

Isn't the article's main concern the conflict of interest when polling companies release both public and private polls?

They have an incentive to release inaccurate results to the public in order to benefit their private clients even further.

[dantillberg]

Polling companies have an incentive already to create inaccurate polls by structuring the questions cleverly. Regulation will never do away with that.

[eftpotrm]

Political polling is a tiny percentage of their business; they mostly do market research for corporates.

Every few years, polls get verified against reality at the ballot box. If you're running a polling company, would you want to sell your market research as the company whose polls were some way off the votes they'd just seen counted because you were trying to influence the electorate?

[Swenrekcah]

I don’t understand how those incentives would work. Could you elaborate?

[mrec]

Pollsters normally have a good idea what answer their customer wants to hear. Those who deliver that answer get more business.

[warkdarrior]

Yes, until their customer sees that the poll results do not match reality.

[mrec]

Polls aren't always about matching reality. They're often used as ammunition in an argument - "Look, the majority supports my position!"

The classic example is https://www.youtube.com/watch?v=G0ZZJXw4MTA

[jonathankoren]

Not every poll is a push poll.

Push polls aren’t even polls. They’re just advertising.

[pessimizer]

There is no definitive difference between a push poll and a poll that is not a push poll. There are shameless push polls, but objective published polls are about as likely as objective public news; I'd need an existence proof for either. Published polls on how people say they will vote have virtually no utility for the vast majority of individuals. A small minority of individuals may need to plan for contingencies based on the likelihood of a particular result, but that utility drops drastically the shorter the planning window gets (i.e. the closer the vote is), which might be somewhat offset by a possibly greater accuracy. But instead of slowing as the date arrives, the polls get faster and faster.

However, the utility of a "push" poll goes up closer to the vote, because voters will have less opportunity to review and correct misimpressions that the poll gave them. Hence the time between polls goes to zero as time to the vote approaches zero, and why some governments feel that they have to step in to create a prohibited window before a vote.

Also, a traditional definition of a "push" poll is one that is meant to change the minds of the people taking the poll, but more effective in many circumstances, and definitely more common (because it is cheaper) is a poll which uses strange phrasing to get results that can be reported (after being carefully framed in a press release) in a way that implies far greater support or opposition for something that would be indicated by a more direct questioning. The fact that every vote is to some extent a Keynesian beauty contest, even if that's simply because people are convinced by ad populum and question themselves if they see that the crowd disagrees with them, means that even an "objective" poll would push.

[jonathankoren]

There’s a lot to unpack here, but you seem to misunderstand why election polls are reported, and reported more frequently the closer it gets to Election Day.

Tracking polls aren’t for informing some secret cabal, or even to inform. It’s the horse race. It sells papers. But there’s nothing biased about “Is the country on the right track, or the wrong track?”