Polling: Unskewed Polls and comments on polling (moved from Harris Thread)

Politics & Elections
281

I’m trying in good faith to figure out where your confusion is coming from and the quote above may be part of it. You are not reading what @bordelond is citing. The GOP is the customer. That is the first part of the story:

From the Hopium link:

Clearly Lara Trump and RNC Chair Whatley, former NC GOP Chair, have turned on some machine as no other state has seen this many right-aligned polls in recent weeks. The game here obviously is to not let it appear that Mark Robinson’s meltdown is causing NC to slip away.

From the first of several links in a subsequent post:

The only problem was that the Peak Insights poll was paid for by the National Republican Senatorial Committee

Rosenberg’s specific claim is the second part of the story:

In the weeks leading up to the 2022 midterms, Democratic strategist Simon Rosenberg was ringing the alarm … snip … "In six major battleground states, more than half the polls conducted in October have been conducted by Republican firms… Basically we can’t trust the data on RealClearPolitics or FiveThirtyEight any longer… it’s essentially Republican propaganda

Not that they flooded the zone in June, but that they flooded it in October, while 538 was on the money in October.

The final part is the claim is that currently these GOP ordered false polls are flooding NC and shifting the aggregated results. Yet the aggregated results in 538 and the NYT are exactly in line with what they would be if all you looked at was the most recent NYT, Marist, and Morning Consult results.

In short, assuming it is true that there has been an attempt to flood the polls by the GOP, he is wrong in his assessment that it skewed the results in the final weeks of the ‘22 cycle or that it is skewing the aggregation numbers in NC today.

He, like the GOP folk who order up these GOP good news polls, thinks that getting “good news” out there, true or not, is good somehow. He wants to believe that Harris is comfortably ahead by polling if not for GOP “fuckery”, and finds a willing audience of people willing to give him clicks to hear that story.

The aggregation numbers are highly likely going to be off … because they usually are, no fuckery required.

282

Would a grifter lie to the customer?

Yes. Because that’s what grifters do.

283

Some people hypothesized that Hillary Clinton was so far ahead in 2016 polls, that many went ahead and voted 3rd party or stayed home.

Is polling that good are generating income?

284

Ask this guy…

285

And that answers my question- how?

286

It was not intended to answer your question. Rather than speculate I directed you to someone who would know the answer.

For future reference, when you ask someone a question and they suggest that you ask someone else, it’s not meant to be an answer. More of a helpful response. You’re welcome.

287

How do I ask a wikipedia link?

Not helpful.

288

Interesting.

I knew they mostly did it to advertise their marketing research business. If claims seen in this thread, that pollsters commonly put their finger on the scale to give desired results, come to be widely believed, it will become bad advertising, and they’ll stop doing political polls. Perhaps university-sponsored polling will take up the slack.

The Harvard Center for American Political Studies (CAPS) - Harris Poll seems to want to get both kinds of marketing bumps.

289

Sort of exactly? These are false polls that push the other direction. Based on a belief that voters want to vote for the winner I guess?

I have no evidence in support of this but my belief is that a belief that the election is close and that “your” candidate is slightly behind and needs your vote and help, that anxiety, drives turnout and funding more than a belief that an election is in the bag. On both sides.

These GOP leaning polls are commission work. The GOP pays for them in a belief that “good news” polls helps them somehow. I suspect they get paid well enough. And they do allow certain pundits to cherry pick them to tell a story that drives clicks. It allowed some media pundits to tell a red wave story that the aggregate data did NOT support. They do impact RCP numbers slightly. They do not seem to impact the actual aggregators’ numbers in any negative way though.

290

Thank you, that does answer my question.

291

The Wikipedia link was to a page about a person. You can ask people questions.

Are these real questions?

292

And so? Does it include contact info? and why ask him?

293

The New Republic has issued an article which states that Rasmussen is secretly providing its polling results to Republicans in advance of release, even though it purports to be non-partisan and operated by a nonprofit organization that is barred from supporting a political party by virtue of its tax status.

538 earlier dropped Rasmussen from its aggregators, because of unclear methodology, for which it was attacked by Republicans.

294

Elliot Morris, who now runs 538, has previously been all in on ignoring certain pollsters, and downweighting others.

He acknowledges that including the “bad” pollsters may have made the result more accurate, but does not like being right for the wrong reasons. It’s a fun read.

I’m assuming that he has implemented changes based on those analyses in the current 538.

295

Rasmussen has gone full election denier since being outed.

296

That’s a weird take. If my job is to make a prediction based on the entire universe of publicly-available data why would I intentionally make my predictions worse by excluding data based on some vague notion of “badness”?

Imagine I’m predicting football scores and I know that one bookmaker just really likes road underdogs so they consistently move their line one point in their favor vs. everybody else. I could say “that’s bad process” and exclude their lines from my averages. But if the data shows that, for some reason, including them produces more accurate predictions than excluding them I really don’t get it.

Or, put another way, is there not a possibility that the explicit biases of the partisan pollsters is somehow helping to balance out the implicit bias of the non-partisan ones? Maybe these “priming” questions, which seem like a terrible way to get an honest response, actually somehow mirror the way voters actually think when they go in the voting booth?

Anyways, I will definitely put more stock on the aggregators that include all pollsters but adjust for house effects and past performance (538, now, and Silver) over those that just average a seemingly random number of polls (RCP) or exclude a large numbers of them (which Morris seems to advocate for in the Substack but doesn’t seem to actually do at 538 now that he is in charge there).

297

His take is that being right for the wrong reason will lead to poorer performance in the future. But yes it a point of philosophy as well, do

we collectively want an analysis of public opinion that is balanced both towards predictive accuracy and correctly measuring and understanding the general will.

He believes that including more garbage impedes the latter.

298

What you’re missing is that predicting football games of which there are hundreds per year is different from predicting presidential races of which there is one every 4 years.

After your model has done 500 presidential elections we’ll talk again.

299

That’s a great point. We just don’t have enough data to know if what the “red wave” pollsters was hackery that happened to be right once or a modified polling technique that generates better results.

300

Re partisan discussion of partisan pollsters, Elissa Slotkin says that a Democratic campaign pollster (unlike most public polling) has Harris losing Michigan:

Axios: Rep. Elissa Slotkin warns Harris is “underwater” in Michigan

Worrisome for Democrats? A little. But I have no reason to believe Slotkin’s polling operation is any better or worse than the average public pollster.