Why do they do national polls?
May 25, 2016 11:13 AM Subscribe
The American public will be inundated with national poll results until November 8th, and I can't understand why.
The USA doesn't have a national election, there are 50 state elections. The national vote count doesn't matter, all that matters are each state's winner, which assign their electors. And more than half of these states aren't really in play.
So, why do national polls? Who cares? (except for Libertarians, who really want to poll 15% so they can be in the debates)
The USA doesn't have a national election, there are 50 state elections. The national vote count doesn't matter, all that matters are each state's winner, which assign their electors. And more than half of these states aren't really in play.
So, why do national polls? Who cares? (except for Libertarians, who really want to poll 15% so they can be in the debates)
The news media like these because it gives them something easy to talk about. This kind of poll is also cheaper than doing 50 per-state polls.
posted by Chocolate Pickle at 11:27 AM on May 25, 2016 [1 favorite]
posted by Chocolate Pickle at 11:27 AM on May 25, 2016 [1 favorite]
Doing a national poll costs less than 1/50th of doing 50 state polls and collating the results to present an electoral college result.
posted by Etrigan at 11:28 AM on May 25, 2016 [2 favorites]
posted by Etrigan at 11:28 AM on May 25, 2016 [2 favorites]
A national poll is much easier to take than 50 state polls, but is still very predictive of the end result. Out of 57 elections, there have been only four instances of a candidate winning the popular vote but losing the election.
posted by ejs at 11:28 AM on May 25, 2016 [8 favorites]
posted by ejs at 11:28 AM on May 25, 2016 [8 favorites]
24-hour news cycle needs "new" content to gnash on constantly.
posted by Lyn Never at 11:32 AM on May 25, 2016 [3 favorites]
posted by Lyn Never at 11:32 AM on May 25, 2016 [3 favorites]
Lots of people care—I think most people who care about this stuff at all in a non-jaded way (i.e. people who watch cable news, etc.) A) think of themselves as Americans more than as residents of a certain state and B) want to (and do) think that their vote matters as something more than a state-level way to elect a proxy-voter later on.
posted by Polycarp at 11:32 AM on May 25, 2016 [1 favorite]
posted by Polycarp at 11:32 AM on May 25, 2016 [1 favorite]
Because it's hard to make someone in Ohio care about what Florida thinks (in a sort of general sense) but if you say "The COUNTRY thinks this..." you can make an information product that you can sell to people in all fifty states.
More jaded answer: we need to maintain the polite fiction that this election isn't basically all about Ohio, Florida and whatever other states emerge as "swing" states.
posted by jessamyn at 11:34 AM on May 25, 2016 [8 favorites]
More jaded answer: we need to maintain the polite fiction that this election isn't basically all about Ohio, Florida and whatever other states emerge as "swing" states.
posted by jessamyn at 11:34 AM on May 25, 2016 [8 favorites]
Polling. It's a job.
posted by IndigoJones at 11:57 AM on May 25, 2016
posted by IndigoJones at 11:57 AM on May 25, 2016
Branding.
National polls can be referenced at all levels of news- local, regional, and national. When they mention your poll, they mention your name, which reinforces your brand as a credible source of poll data.
posted by mkultra at 12:04 PM on May 25, 2016 [2 favorites]
National polls can be referenced at all levels of news- local, regional, and national. When they mention your poll, they mention your name, which reinforces your brand as a credible source of poll data.
posted by mkultra at 12:04 PM on May 25, 2016 [2 favorites]
FiveThirtyEight addressed this recently. Basically, as ejs says the national poll is usually predictive (unless its very close, then you know you might have to pay more attention to state by state stuff).
"There’s about a 25 percent chance of a split if the national popular vote is decided by about 1 percentage point, and that the chance is cut in half when the margin is 2 percentage points."
So once you get beyond a couple percent difference, the chance of a split is extremely low and national polls should give you a good idea.
They also point out that because state polling is sparse, you shouldn't trust it as much (of course, this could be fixed by doing a ton of state polling, but that seems unlikely to happen).
posted by thefoxgod at 12:08 PM on May 25, 2016 [2 favorites]
"There’s about a 25 percent chance of a split if the national popular vote is decided by about 1 percentage point, and that the chance is cut in half when the margin is 2 percentage points."
So once you get beyond a couple percent difference, the chance of a split is extremely low and national polls should give you a good idea.
They also point out that because state polling is sparse, you shouldn't trust it as much (of course, this could be fixed by doing a ton of state polling, but that seems unlikely to happen).
posted by thefoxgod at 12:08 PM on May 25, 2016 [2 favorites]
As others have pointed out, a national poll is 1/50th the cost of polling every state, for much much more than 1/50th the predictive power.
But even more importantly, the cases where the national/state distinction matters are the cases where no poll is going to give you a definitive answer. Given that states have a reasonably strong tendency to shift in tandem, for a electoral college/popular vote split to occur, you need a popular vote margin within one or two points, which implies that several states will be within one or two points. That means that the election is well within the margin of error either way, so why spend a lot of extra money to essentially say "We know with very high certainty that we don't know who will win"?
posted by firechicago at 12:35 PM on May 25, 2016 [2 favorites]
But even more importantly, the cases where the national/state distinction matters are the cases where no poll is going to give you a definitive answer. Given that states have a reasonably strong tendency to shift in tandem, for a electoral college/popular vote split to occur, you need a popular vote margin within one or two points, which implies that several states will be within one or two points. That means that the election is well within the margin of error either way, so why spend a lot of extra money to essentially say "We know with very high certainty that we don't know who will win"?
posted by firechicago at 12:35 PM on May 25, 2016 [2 favorites]
Also, no polling organization worth its salt (or budget) asks only one question on national polls. They ask a lot of questions, about all kinds of things, depending on a variety of factors.
posted by rtha at 12:50 PM on May 25, 2016 [1 favorite]
posted by rtha at 12:50 PM on May 25, 2016 [1 favorite]
In Canada, at least, election polling is also used by pollsters to raise their profile for non-election opinion research. If you've spent the last ten months hearing about Pollster Inc. polls in the news, maybe they're the people you call when you need to know what women aged 22-36 are looking for in a car (and you're willing to pay to find out).
posted by Clandestine Outlawry at 1:16 PM on May 25, 2016 [2 favorites]
posted by Clandestine Outlawry at 1:16 PM on May 25, 2016 [2 favorites]
This thread is closed to new comments.
posted by notned at 11:26 AM on May 25, 2016 [4 favorites]