Steady Data, Shaky Narratives: Why Poll Hype Misleads & Slow Drift Rules in Understanding "Public Opinion"
G. Elliott Morris deservedly trashes the once-reliable Harry Enten for feeding the ad-supported eyeball-gluing misinformation beast. & it leads me to harken back to the true Glory Days of Nate Silver…
G. Elliott Morris deservedly trashes the once-reliable Harry Enten for feeding the ad-supported eyeball-gluing misinformation beast. & it leads me to harken back to the true Glory Days of Nate Silver…
The wise G. Elliott Morris is very unhappy with CNN’s Harry Enten.
Why? Enten is now deceiving his viewers on TV in order to serve their ad-supported eyeball-glueing mendacity-serving business model:
G. Elliott Morris: Is Trump “doing what he promised”? <https://www.gelliottmorris.com/p/is-trump-doing-what-he-promised-weekly>: ‘I don’t normally devote the lead slot in this weekly roundup just to disagreeing… but I’m going to make an exception…. Harry Enten claiming Trump is “doing what he promised” and “might as well be called Steady Eddie” in the eyes of the public… that public opinion about Trump has not changed since 2024, and later implies he could win again if a hypothetical do-over happened today…. The three most relevant quotes…. “Go back to October of 2024, 44 percent of the American public viewed him favorably. Now, despite everything, look at this, 43 percent…”. “The race for Congress, Dems versus Republican margin…. It was plus three Democrats at this point a year ago. Now it’s plus four…. We had [this] a year ago, and it was not good enough for Democrats to take control…”. ”52 percent, the majority of Americans say that Donald Trump is doing what he promised…. Trump has basically the steadiest favorable rating this much through a presidency of any president on record… Good enough to get him reelected…”…
Two things going wrong…. Cherry-picking… results… almost entirely driven by which polls he’s selecting, the questions he’s using, and the timeline used for his comparison.… [For example,] the president’s job approval rating—a number which is more predictive than favorability of things we care about, like midterm and presidential election outcomes—is significantly worse than it was at the start of Trump’s term….
[And] our average of polls at 50+1 shows a 7-point swing away from Republicans on the generic ballot since the start of Trump’s term, when the GOP was up 3 vs Democrats….
Yes, 52% of Americans say they think Trump is “doing what he promised.” But Enten implies that this is somehow a rosy figure for “Steady Eddie” Trump…. [But[ many people think Trump is “doing what he promised,” and they do not like it….
Since the directional informational content of the question is low… the implied 50% benchmark is nonsensical….
On any given day, according to Enten, the sky is either falling for Trump or he’s destroying the competition…. Sensationaliz[ing] findings to keep viewers engaged… baked into the business model…. But… with public opinion… changes are slow and the story is pretty obvious, and often boring. A segment that says Trump has been slowly losing ground since the start of his presidency, and Republicans are probably heading for a midterm loss, is not going to hold an audience for long…
I think Morris is 100% right here.
And I am very happy he is doing this.
And I am especially disappointed to see this from Harry Enten, whose “Margin of Error” blog <https://marginoferror.org/> was quite good, and who did impressive work at Nate Silver’s FiveThirtyEight <http://fivethirtyeight.com> <https://abcnews.go.com/538> < https://www.natesilver.net> back before 2018.
So let me harken back to The Day. Remember: Back when Nate Silver showed up, it was a complete and total breath of fresh air. Why? Precisely because Nate did not do the excited IMPORTANT BREAKING NEWS NEWS!! panic-octopus embarrassed-penguin parade dumpster salsa dance-dance.
What did he do instead? Nate Silver reported polling averages. They showed public opinion today very much like yesterday, and like two weeks ago, with perhaps a slow drift one way or the other over months. He did the unsexy thing: he let the numbers speak for themselves. That said that sentiment moves glacially, not in cable-news lurches. That was the point. Nate Silver had the discipline and the honesty to foreground the central tendency and the uncertainty around it, rather than to cherry-pick outliers and confect drama.
Nate Silver is and will always be an enormous public benefactor, if only for having stunningly successfully called bullshit on a deeply broken and mendacious political polling and forecasting industry, and then done something about it. I say this, I say this again. And I say this yet again. And again. And I say most of his good work continues.
Now Back in the Day Nate Silver did add an amplification swing-inducing factor to his writings: he showed not just raw averages but modeled win probabilities, derived from those averages and from priors about how votes translate into seats and outcomes. Those can and do show considerable day-to-day variation.
But that was a perfectly legitimate amp: it was an honest evaluation of something we are very interested in for good reasons. And even there Silver’s emphasis was on epistemic humility. When polling margins were close, he refused the crutch of “neck‑and‑neck.” He said our knowledge was uncertain. He said, in effect: the odds that the eventual winner will win could be as skewed as 90–10, but we do not, at present, possess enough information to know which side is sitting in the 90. Thus shifts in win probabilities were, often, much better viewed not as a race horse pulling in front by half a length, but rather a reduction in our uncertainty over which candidate truly had a subtantial lead.
That is what it looks like when you treat inference as inference, not as entertainment.
And the audience he cultivated—those who wanted their brains cleaned of mendacious ad‑supported slop—could then follow deep dives into the other things he reported. He told you what the aggregates were actually telling us: how house effects, sampling frames, and late movement mattered. He told us how the boring story of slow drift is ultimately the real story.
How then did Nate Silver make his nut back in the day? There turned out to be a substantial number who actually wanted to have their brains cleaned of the mendacious ad-supported slop. That was one part of it.
It wasn’t sorcery; it was discipline. A method like FiveThirtyEight’s weighted averages and transparent model diagnostics mattered: it taught readers how to separate noise from information, and it did so consistently across cycles and domains, from elections to sports, building trust through epistemic humility rather than theatrics. The public sphere improves when we privilege aggregates and uncertainty over hot takes; the Digital Commons grows when readers demand it.
Deep dives into the details and the meaning of interesting particular aspects of what the aggregate of polls was actually telling us. The boring-but-crucial lesson of aggregation is that pooled polling—properly weighted for house effects, mode (live phone versus online panels), sample design, and field dates—reveals slow, coherent drift rather than melodramatic whipsaws, and that outliers nearly always vanish back into the mean once you account for method. Consider how 2012’s state-by-state averages, after adjusting Gallup’s house effect and late-sample quirks elsewhere, yielded a stable Obama edge that culminated in all 50 states called correctly; or how 2018’s national swing toward Democrats translated unevenly into seats because district geography and incumbency mattered more than any single “shock” poll; or how 2022’s late-break “red wave” narrative dissolved when rolling averages, not cherry-picked internals, showed only modest movement.
The mechanics do the work: pollster ratings tame systematic bias; rolling windows smooth transient noise; likely-voter screens and turnout models explain why a one-point national shift can mean ten seats in one chamber and two in the other. Electorates evolve on the cadence of months and institutions, not minutes and segments. Aggregation—plus transparent uncertainty—protects us from mistaking sampling error and partisan nonresponse for turning points; it’s the difference between inference and entertainment.
It was a wonderful addition to humanity’s collective intelligence: less dopamine, more signal; fewer “BREAKING!!!” segments, more understanding of how democracies evolve in the data. Plus Nate Silver’s book The Signal and the Noise was, I thought, brilliant. I am forever deeply in his debt for his making me better-informed and wiser across a whole bunch of different dimensions.
Now it is no secret that I think Nate Silver’s VAR has gone down since those days.
But do not get me wrong with respect to Nate Silver and company. He is still very valuable in his data-connected reality-grounded sphere. I look forward to learning a huge amount from his Silver Bulletin over the next couple of years. I think he is still a very good complement to G. Elliott Morris.
And I thought that chapters 1-4 of his new book On the Edge were brilliant. Paul Seabright nailed why:
Paul Seabright: O Lucky Man! <https://www.the-tls.com/politics-society/social-cultural-studies/on-the-edge-nate-silver-book-review-paul-seabright>: : ‘On the Edge is really two books…. The first… [of] a little over 200 pages, takes us on a tour of the world of professional gamblers…. This is one of the most gripping accounts I have ever read of a milieu it would bore me rigid to inhabit myself…. Repeated success therefore requires the ability to calculate probabilities accurately, but also the nerve to make large stakes and to absorb potentially heavy losses along the way…. Poker obviously requires acute skills of social perception…. Sports gambling… involves a subtly different set of skills…. Those who make a living from gambling over long periods are skilled in ways that set them apart from other people… the mindset, lifestyle and other aspects of the personality of those who can reliably assess probabilities to the degree of precision required to outwit each other and the house… more than a whiff of books and films such as The Hustler, The Gambler, Casino, and The Cincinnati Kid…
There is a very good book about the professional gambler psyche trapped inside On the Edge. Had I been Nate Silver’s editor/publisher, I would have sawn the book down to its first four chapters, the ones on poker and sports betting. I would have put bookend chapters around those four—a new “why poker and sports betting matter as windows into our society” introduction and a “what we learn from them about broader issues of managing risk and dealing with counterparties” conclusion. I would have published it as a short thing to read on airplanes. And then, I think, very good reviews and, hopefully, the money would have rolled in.
But the rest of the book is not at all good, in my opinion. Again, Paul Seabright nailed why:
Paul Seabright: O Lucky Man! <https://www.the-tls.com/politics-society/social-cultural-studies/on-the-edge-nate-silver-book-review-paul-seabright>: ‘On the Edge is really two books…. Things take a stranger turn when Silver starts to write of such [successful gambler] people as belonging to… “the River…. A sprawling ecosystem of like-minded people that includes everyone from low-stakes poker pros just trying to grind out a living to crypto kings and venture-capital billionaires”. “Sprawling” seems about right, “like-minded” less so…. Silver tells us that such people share a “cognitive cluster”… analytical and abstract thinking… proficiency in… “decoupling”… the ability to focus on logic and to block out context… competitive, critical of others, independent-minded (even contrarian) and highly open to risk. Professional gamblers belong… but so do Peter Thiel and Elon Musk, to whom many admiring pages will later be devoted… <https://www.the-tls.co.uk/politics-society/social-cultural-studies/on-the-edge-nate-silver-book-review-paul-seabright>
In my view, Silver’s “The River” is made up not of one common mind set, but of six different groups of people. Call them: Gamblers, Speculators, Calculators, Customers, Suckers, and Grifters. When Nate Silver writes:
activities… [of] capital-G Gambling—like blackjack and slots and horse racing and lotteries and poker and sports betting—are fundamentally not that different from trading stock options or crypto tokens, or investing in new tech startups…
I want to say: yes they are that different. Behind investing in new tech startups are customers whose production, entrepreneurship, and enterprise is enabled by the system—an activity with a very large positive-sum component. Behind trading stock options is, sometimes, also an activity in which the counterparty is a customer who has very good reason to lay off risk, and whose well-being and that of his other counterparties is greatly enhanced thereby. Other times not. And with blackjack and slots and horse racing and lotteries and sports betting—well, you had better hope that the counterparties to the calculators are simply gamblers, and that there are few grifters in the mix. For when these activities become, as they so often do, the playgrounds of grifters and speculators and suckers, it becomes a real shitshow. But Nate Silver sees none of this. And that is what sends the book On the Edge, starting after chapter 4, badly astray. Basically, Silver runs into various lucky speculator-grifters, and they successfully make him the sucker in their con games. They do not take his money (I think). But they do take his time and his mind. And he becomes a shill for crypto and AI grifters, including the weird subbranch of those who call themselves “Effective Altruists” who either think they are rapidly building a benificent or fear someone is building a malevolent Digital God.
Let me again quote Paul Seabright, and give him the next tolast word:
Paul Seabright: O Lucky Man! <https://www.the-tls.com/politics-society/social-cultural-studies/on-the-edge-nate-silver-book-review-paul-seabright>: ‘Therein lies the most important difference between such [speculator-grifter] people [in the second half] and the professional gamblers who populate the first half of On the Edge. Professional gamblers are constantly having to correct their methods and their conclusions against an unforgiving daily reality. Crypto kings and venture capital billionaires, in contrast, are continually trying to sell us their wisdom in novel domains, whether it’s by asking us to accept their guesses about the existential risks of artificial intelligence or by accepting the invitation to become vice-president of the US. Such novel domains lack an equivalently rigorous benchmarking against reality. And Nate Silver’s book, which I in many ways like and admire, risks giving unearned credence to this dangerous bait-and-switch…
The lesson I derive from the second half of On the Edge is how dangerously prone to brainwashing for purposes of social agreement our brains are. That makes it really, really important to double down on the good aspects of the work of Nate Silver and others: to stay tied to honest analyses of the data as completely and rigorously as we can. And this is why I think it is really a very bad idea for people like Harry Enten to say things that are not right.
References:
Byers, Dylan. 2012. “Nate Silver: One-Term Celebrity?” Politico, October 29. <https://www.politico.com/blogs/media/2012/10/nate-silver-one-term-celebrity-147618>.
Confessore, Nick. 2017. DELETED. Twitter. <https://twitter.com/nickconfessore/status/846186757909549056>.
DeLong, J. Bradford. 2024. “Two Cheers for Nate Silver & Co.: Polls, Predictions, & the Persistent Problem of Political Noise”. Grasping Reality. October 8. <https://braddelong.substack.com/p/two-cheers-for-nate-silver-and-co>.
DeLong, J. Bradford. 2024. <https://braddelong.substack.com/p/time-to-write-my-negative-review>.
Enten, Harry J. n.d. Margin of Σrror. <https://marginoferror.org/>.
Morris, G. Elliott. 2025. <https://www.gelliottmorris.com/p/is-trump-doing-what-he-promised-weekly>
Poundstone, William. 2005. Fortune’s Formula: The Untold Story of the Scientific Betting System That Beat the Casinos & Wall Street. New York: Hill and Wang. <https://archive.org/details/fortunesformulau00poun>.
Robinson, Nathan. 2017. “Why You Should Never, Ever Listen To Nate Silver”. Current Affairs. December. <https://www.currentaffairs.org/2017/12/why-you-should-never-ever-listen-to-nate-silver>.
Seabright, Paul. 2024. “O Lucky Man!” Times Literary Supplement. September 2. <https://www.the-tls.co.uk/politics-society/social-cultural-studies/on-the-edge-nate-silver-book-review-paul-seabright>
Silver, Nate. n.d. Silver Bulletin. <https://www.natesilver.net>.
Silver, Nate. 2024. On the Edge: The Art of Risking Everything. New York: Penguin Press. <https://www.google.com/books/edition/On_the_Edge/doPqEAAAQBAJ>.
Silver, Nate. 2012. The Signal & the Noise: Why So Many Predictions Fail—But Some Don’t. New York: Penguin Press. <https://archive.org/details/signalnoisewhysa0000silv>.
Silver, Nate. n.d. FiveThirtyEight. <http://fivethirtyeight.com>.






Brad, thanks for that and by the way your stacks are out of the world good. Substack would be much much less interesting and complete without them.
Brilliant Brad, right on the money. At the end of the day, all cable News show succumb to temptation to make everything Breaking News to keep the eyeballs glued and rolling