Why is it treated as a given "the polls always tighten"?

[Silent3]

If it's true, it's true, I won't argue against the facts if that's what the recorded history of polling shows.

But it doesn't make a lot of sense to me.

In fact, it would make more sense to me if impressions of candidates hardened as time went on, and leads tended to freeze or expand (barring significant "October surprises" ).

Even if a lot of people are undecided until elections draw closer, in order for races to tighten, that would have to mean that for some strange reasons undecided people are more likely to go for whoever started out behind.

Is the issue something in polling models themselves, and some changing criteria for who counts as a "likely voter" as time runs down approaching an election?

[AZSkiffyGeek]

1. There is a switch from "registered" to "likely"

That is probably part of it.

[Polybius]

2. Bill Clinton just said something similar

If anyone here has a link to it, please post. Republicans are also notorious for October surprises.

[Greybnk48]

3. They can say whatever they want (the media), how would we know different?

I think they say whatever they think will keep people glaring at the screen.

[Silent3]

4. Polling firms aren't conspiracies

Well, maybe a few are just making shit up, but for the biggest pollsters political polling is a form of advertising for market research, meant to demonstrate the quality of the company's research to potential paying customers.

Fudging polls for the sake of creating drama might help a broadcaster, but it's terrible for the pollsters themselves, who can only harm their reputations and billing rates if they produce garbage numbers that don't reflect reality as closely as they can manage.

[Ohio Joe]

5. Just a guess but...

There are not a lot of huge blowouts, races are often within a few percentage points. As it gets closer to the election, more of the undecided voters make up their mind, so unless the race is going to be a blowout, it should tighten. At least that makes sense to me.

[Silent3]

6. But why should undecided voters favor the running-behind candidate more?

Voters who finally decide who they are voting for don't replace already-decided voters, they only add to their numbers.

For the shift to consistently leading to tightening, the newly-decided would have to always break opposite from the previously-decided.

[Ohio Joe]

7. Still guessing...

And definitely anecdotal but... I known a lot of people over the years who always vote the same but until a few weeks before the election they always say they are not sure who they will vote for yet.

I've got a buddy out in MA that I've known since kindergarten and he has this weird way of voting. He always votes against the incumbent no matter what, but he'll always say he is not sure up until election day. I've even called him on it a few times and he just goes... Oh yeah, I do do that don't I. It's pretty funny.

This is all just guessing on my part though, I know almost nothing about polling.

[TheBeam19]

8. Could be because the incumbent is

a known quantity (in terms of job performance in the job being sought), and if an undecided voter is on the fence but can’t be convinced to stay the course with what they already know, then they’ll break for the challenger toward the end of the campaign. Sort of a default choice.

Just a guess. And of course doesn’t apply to races in which there’s an open “seat.”

[doc03]

9. It's to make it a horse race for ratings and more

money.

[Silent3]

10. That presumes polls are just made-up numbers

What evidence do you have of that?