Dang it, my Pixel phone’s touchscreen is the absolute worst.
But… Thanks for the bunched Corgis
5 Likes
Couldn’t resist, the image was just too good.
7 Likes
Silver is allowed to be wrong, allowed to be fooled by bad polls, allowed to jump the wrong way as to whether to include a new poll, allowed to make a model based on the last election that fails to perfectly model the next election.
All of these things are fine (if not the desired result). None of them disqualify him, or any poll guru.
The problem comes when—as he has done four major elections in a row now—he reacts to people noticing that these things happen, even if it’s not phrased as a criticism, with a sniffy little tantrum about he did not fail, but WAS failed. The model is fine, you jerks, it’s just that you’re not using it right.
If you were watching 538 in 2018, the model ingested the very first results for rural districts on the east coast. It correctly predicted that the GOP would easily hold these deep red seats. Then it looked for Democratic victories to offset them and found none, because they hadn’t been reported yet, so it started trumpeting a massive pro-Trump upset. Silver sent a lesser Nate out to shrug about it, and then for weeks afterwards defended himself by saying that it eventually started predicting reality correctly once reality had already happened. [slow clap]
8 Likes
You mean the young kids who stood in line for hours because their rights to bodily autonomy were being taken away? Those lazy assholes? Or the Black women who saved democracy in Georgia? Those lazy assholes? I think you know not of what you speak.
27 Likes
He’s become quite the gambler, too, participating in a lot of poker tourneys in Vegas. I think it’s clouding his judgement.
3 Likes
I stopped using 538 as my bellweather when he was so very very wrong about 2016 and then complained that we read his data wrong.
6 Likes
No, I mean the people 6 years ago who assumed that Clinton would win so just didn’t bother to vote.
Tell me that if the polls were showing a Democratic landslide those people would come out to vote?
2 Likes
I don’t get this attitude. How can you (not you personally, society as a whole) complain about polls being inaccurate but also be willing to lie to them? It’s perfectly fine to decline being part of a poll but once you’re in it you should answer it truthfully. Otherwise we just all lose.
ETA: Yeah, I’m basically demonstrating what @fnordius wrote in the other election thread:
6 Likes
Maybe rather than complaining about the lack of balance in partisan polling on the left, he should be seeking better polls?
8 Likes
I’m not so sure that better polls exist. There is obviously a growing gap between the kinds of people who answer pollsters and the kinds of people who actually go out and vote, and I don’t think that there is any easy way to calibrate polls to account for “people who don’t want to answer pollsters but still want to vote.”
Random sampling doesn’t work if the sample is already stacked in favor of people who are inclined to answer a poll and such people are not reflective of the electorate as a whole.
12 Likes
I don’t think that is a bad question. I used to take surveys and polls seriously but I got a few of polls in late-90s to the mid-2000s where that literally omitted the strong agreement and disagreement on key questions that weren’t omitted on others. It was clear that the surveys were designed to slant the data in one way.
I remember one survey question was about Wes Clark. The question was something like, “Considering that Wes Clark is an actual military veteran and Howard Dean and John Edwards are not, How likely are you to support Wes Clark’s candidacy for president? 1 - Very Likely, 2 - Somewhat Likely, 3 - Undecided.”
I replied, “4 - Somewhat unlikely.” After a beat, the operator says, “That’s not an option. It’s 1, 2, or 3.” I’m just confused by this, so I probe. “I can’t answer your question in the negative?” He replies, “I don’t write the surveys.”
There was another survey about local politics where they asked about giving tax cuts or breaks as an incentive to retain or attract business to my area. I could not answer anything that suggested I was against it. Again, the survey would not let me express disagreement with policy idea even though other questions were on the 5 point scale. I could be undecided or express two levels of support and nothing more.
So, I began outright lying to pollsters. There is no benefit to constructively engage with pollsters who craft their questions to create pre-ordained conclusion.
Every political poll should feature yes / no questions, three stage questions (agree, disagree, neither), five stage questions (SA, A, NAoD, D, SD), or strict choice (which candidate to you favor). Anything else (especially 4 stage SA, A, D, SD without NAoD or three stage without an agree or disagree option is manipulation.
13 Likes
What if “reputational risk” is just something he’s making up out of thin air and doesn’t exist? It’s not like everyone including the those GOP biased pollsters doesn’t already know they are biased. Which makes that a zero cost thing. Everyone expects the GOP to be biased and lie, they expect better from the DEM, it’s not an equal comparison, even if it did mean something.
That part seems like a huge self own. He’s literally saying the model accounts for the effect, except it was wrong anyway since the other side didn’t also create fake data. So, which is it? If his model accounts for garbage, why was it wrong? And, why would the other side submitting different garbage change the model results if it is accounting for it? It reads like an admission that the model is supposed to account for the garbage, but that it doesn’t.
6 Likes
That’s the question I’m left with: why include the most obvious and reeking garbage at all?
8 Likes
Where and when did he say his model was wrong? I’ve seen him say just the opposite.
It’s not like the end results of this election appear to be dramatically different from 538’s final projections.
2 Likes
Then why not just tell them that they are clearly trying to manufacture an outcome and that you will no longer participate in the poll?
If they do push polling in the first place, they don’t care that you are lying to them, they will get their desired result anyway. Just walk away in that case. Then at least you’re increasing their costs or reducing their sample size, rather than giving them what they want.
6 Likes
This right here.
Republicans see an advantage in sharing polls that show they are in good shape. Their base is going to vote regardless, and their more swingy, low information voters rely on media perception of “momentum” and often subscribe to a “well, seems like most other people like this guy so I guess I’ll vote for him” last minute decision style.
Conversely, Democrats see little advantage in sharing a poll that shows they are in good shape. Their base is motivated by a sense of being close but behind, and they worry that any sense the election is “in the bag” will cause many of their lower commitment supporters to stay home — basically fearing a repeat of 2016.
So while Silver’s “solution” would help him, it’s not very likely because it would not help Democrats, who if anything may internally cheer the release of these rightward biased polls as it helps motivate their base.
2 Likes
s/children/data
It was no secret that the GOP was flooding the market with biased polls. That Silver didn’t update his model ahead of time to account for it says something. 
17 Likes
I think if Silver admitted that the poll zone was flooded with garbage, he’d be out of a job. I’m reminded of my old statistics classes about systemic errors and feel that he should just find another line of work (like go back to quietly making money betting on sports). But then he wouldn’t get to be important and won’t be invited on all the political podcasts and news shows any more… No way his ego could take such a hit. So he puts the garbage through his beautiful model and tries to make things work.
Nothing really useful comes of it, but he still can claim all was within the bounds of error so it didn’t exactly fail. I wonder how much longer the media will fall for this grift. Could be a while. Anyway, he’s put the day of reckoning for his ego off for at least another two years, so good for him (not so good for the rest of us, though).
8 Likes
You should have left it as bunches, because that perfectly illustrates one of the problems with polls and surveys: ambiguously worded questions. The question you asked originally would result in data that’s complete garbage even if the question were presented to the public along with the results.
I have not trusted polling numbers since the 2016 election. I actually think Democrats probably would have done even better in this election if all the media hadn’t been so busy the last six months predicting a red wave and doing their best to make it happen. Polls have changed from being a tool to inform a candidate how well their message is being received to being a tool with which to manipulate voters.
4 Likes
Perhaps “wrong” is to strong. The BB write up said “Nate Silver feels his work would have been more accurate”, which I read as “wrong”.
4 Likes
