Sunday, October 23, 2016

More curiosities surrounding Reuters-Ipsos polling

I'm not able to make the time to do the digging I'd like to at the moment, but playing around with the R-I interface that allows users to toggle cross-tabs, I noticed that since September 1st, R-I has sampled just 376 likely voters in Texas to 565 likely voters in Virginia. The 50 states and DC add up to the entire nationwide sample so there isn't any weighting of responses taking place as R-I gets to its national polling numbers.

Given that Texans cast more than twice as many votes in the 2012 presidential election as Virginians did, that's a bit... odd. And given that Texas is a red state while Virginia is a blueish-purple one, maybe it's actually more Oddjob than just odd. There's more than a little evidence that the deck is being stacked.


Stiltdancer said...

Trump doesn't have a gender gap. He has a marriage gap.

Likely Voters
Married Women: Trump - 41.5%, Hillary 36.9%
Single (never married men): Trump - 50.6%, Hillary 32.5%
Single (never married women): Trump - 59.2%, Hillary 18.5%

James said...

You might want to see
Boom! National Poll: Trump Leading 67% To Hillary’s 19% Of 50,000 Voters
I suppose Anonymous might be biased, but they have sure tried to beat the crap out of trump over the last year.

Audacious Epigone said...




That's not based on a scientific sample, unfortunately.

Dan said...

Today's Podesta leak included strategies designed to skew polls.

The dishonesty is maddening.

Sam said...

Is there really that much evidence of a reverse Bradley effect in this election?

Based on the primaries it would seem the opposite:

At this point this seems to be the only way Trump can win it bar the unforeseen (terror attack, explosive e-mail,etc.) but not sure this is all too different from Republicans in 2012 hawking polls showing Romney winning. The only difference I can imagine is that primary crowd isn't sufficient representation of"hidden" votes popping up for the general.

Dan said...

This blog found Trump outperformed polling expectations in the primaries.

Audacious Epigone said...


He underperformed early and overperformed later. Taking a mean without adjusting for population (and that post does) is meaningless. Trump way underperformed in caucuses but they aren't representative of anything at all. If we just look at primaries, Trump did better than expected and if we look at primaries after Super Tuesday, he did way better than expected

That said, I have no idea how that will transfer to the general election (if it does at all).

The Z Blog said...

The trouble with polling is that it only takes some hint of shenanigans to cast the whole enterprise in doubt. A poll showing Clinton winning by 12 points, for example, is plausible only if it is consistent with other polls. When the big national polls show a tie with a healthy and fluid undecided vote, a 12-point margin says the poll is trash and it should never have been published. The fact that it is rushed out as legitimate leads to skepticism about all polling. After all, if you will publish that garbage, you will fake anything.

My theory for a while is that it is not so much chicanery at work as confusion. The old models are not proving useful. Polling outfits cannot admit that as it invalidates their reason to exist. The natural response is to huddle around the coin flip range. They keep tweaking their models and sampling to get closer to that comfortable zone of a tie. All of them seem to be drifting to that happy place. If all the pollsters declare the race to fluid to call, then no one gets blamed.

Random Dude on the Internet said...

Caucuses are trash and were hostile to new voters who weren't aware of what was going on and awarded/benefited machine politicians like Cruz supporters. Trump always did great in primaries where it was just filling out a ballot and turning it in. If anyone wants to argue that Trump will do poorly because he fared poorly in caucuses, then I'd just stop reading then and there. Trump did poorly initially because the first couple of months had most of the caucuses. The latter two months were almost all primaries, where he did exceptionally well.

If there is one thing to be concerned about, it is that the famed "monster vote" will be a no-show or significantly overestimated. The people who can best be described as low information low motivation white voters. They've been so alienated from the process that they decide it's not going to matter. Either they think Trump will win so they don't need to vote or they think Trump will lose so why bother showing up. They are Trump supporters, they might even attend a rally or two, but they figure they're not going to influence the outcome so they will just stay home. If you know any of these people, it's important to get them to register and go vote.

Audacious Epigone said...


ABC got right on that, dropping from Hillary +12 to Hillary +5 in less than a week!

Random Dude,

That people would drive hours and then stand in line for hours to attend a rally but not take the 15 minutes required to vote on election day (or in the weeks leading up to it) is, non-ironically, deplorable. Let's hope that's not who we are in this, the current year!