A better use of polls
People usually consume polls in a passive manner, as they would a sports game or a television program. In Friday’s Washington Post I argue (gift link) that in a close election year, we can use polls to guide our canvassing and donations – using Bayesian reasoning to estimate where we can make the most difference.
Open thread.
Topics:
Dr. Wang–
Might you take a moment to opine on the recent cliff-dive the electoral college median result has performed in the last few days. Other sites note / claim that there has been a slew of biased or intentionally misleading polls over the last week, basically designed to sow uncertainty and plant seeds for future allegations of rigging. No one (that I have seen, anyway) has produced an actual list of such polls and their funders, but it has been discussed as a truth even in reputable / mainstream sites. Do you note such a process in your incoming data, or account for biases in polls, or is it more of an automatic throw-it-in-the-hopper process ?
Thank you.
–Curious John
Agreed, I’d very much like to hear your insights on the current state of the model.
While I agree that the latest aggregations of polls are alarming, I think it’s worth noting the irony of these questions leading the comment section of a post about productive alternatives to fretting about polls. As Sam’s WaPost article notes, this is a time of year when people will “… sit with their devices and obsess over the margins in the latest swing-state polls” but there are easy routes for all Americans to influence the surprisingly high percentage of other Americans who have urgent reasons to apply their votes to knife-edge questions at the polls.
I think it’s OK to ask. I haven’t left another place to do that, have I? 😉
The tracker is currently set to last-one-week or last-three-polls per state, whichever is more data. The window is set short in anticipation that data will become more abundant as the election draws closer. If I hand-calculate the effects of widening the window to last-two-weeks, it doesn’t change the picture much.
Basically I think that there’s no action item at present. The red “strike zone” straddles both sides of 270 electoral votes, and there are three big questions that could go either way: the Presidency, the Senate, and the House. Add in ten legislatures on a knife edge, ten states with abortion on the ballot, and at least nine states with democracy-reform measures, and you can see how fluctuations in the Presidential tracker are just one item on my mental dashboard.
You are all encouraged to go look at votemaximizer.org to see where you can make a difference with your canvassing and your donations. 80% of the U.S. population lives in a place where there’s something of note on the federal or state ballot.
Not exactly fretting.
I live in one of the most scarlet of states so I know precisely how much my vote–or my personal involvement with local politicking–will impact things. If anything, doing the latter will get my tires slashed.
It’s simply a large and unexplained change to a result, from a mechanism that is supposed to work a bit differently from standard poll aggregation.
And where would you have me put it ? On last month’s article, where no one will see it ?
Can you address what Simon Rosenberg is saying about all these right wing groups flooding the 538 polling averages with fake-red-skewed polls? https://youtu.be/IIGdI2HKhqw?si=8t-gSwyzcAwzj7II
Dave, I’d rather not go down the rabbit hole of looking at individual polls. It feels like overkill, which is the way I feel about most detailed corrections in this domain of work. If you can show me that the 2-week poll median is different when one excludes *all* partisan pollsters, then we could get into it.
I believe the in-house analytics algorithms at FiveThirtyEight deal with your concern. They write about it here: “We apply a partisanship adjustment…Our initial guess is partisan polls overestimate support for the allied candidate by 2 percentage points, but this effect will shrink (or grow) if the data reveals a different answer.”
Anyway, with the amount of systematic uncertainty, any correction is going to be a fraction of a percentage point at most. That means the red zone is going to stay on both sides of 270 EV. So it’s still a jump ball no matter what.
Hi me again from races past. Jack Rems & I are Dark Carnival Bookstore.. You may remember me sending you a flying pig. I see Curious John’s question about slanted polls & your statements about limiting the number of polls examined. But I reman curious if the uh poll scene is largely a right-biased mess like Politics Girl soothingly suggests in her 4 minute don’t panic video https://youtu.be/k1VThfAs1ww?si=_RX_cyZwRRZMuZ99
I do see the last few days here skewing minutely to the right.
My question (which admittedly is way below your pay grade) is (1) does Politics Girl make a legit point even for understanding the Highlights on PEC ?
Also (2) does the big Georgia turnout seem to you like good news for Democracy?
And (3,4) Which month do you expect the final result of this election to be known? Is the final possibility really in March?
Thank you so much for being there. I’ve given more than I can to ActBlue via their 800% deals. (5) Are those truly bogus?
Polls seem OK to me, and I don’t notice enough of a skew to matter for one’s advocacy.
The quality and advisability of any ActBlue (or WinRed) fundraising will depend on the quality of the information that is brought to the table. As you may be aware, I provide information on these platforms that is driven by data analytics. See my recent Washington Post piece.