Key excerpt (But it's worth reading the full thing):
But the real value-add of the model is not just in calculating who’s ahead in the polling average. Rather, it’s in understanding the uncertainties in the data: how accurate polls are in practice, and how these errors are correlated between the states. The final margins on Tuesday were actually quite close to the polling averages in the swing states, though less so in blue states, as I’ll discuss in a moment. But this was more or less a textbook illustration of the normal-sized polling error that we frequently wrote about [paid only; basically says that the polling errors could be correlated be correlated between states]. When polls miss low on Trump in one key state, they probably also will in most or all of the others.
In fact, because polling errors are highly correlated between states — and because Trump was ahead in 5 of the 7 swing states anyway — a Trump sweep of the swing states was actually our most common scenario, occurring in 20 percent of simulations. Following the same logic, the second most common outcome, happening 14 percent of the time, was a Harris swing state sweep.6
[Interactive table]
Relatedly, the final Electoral College tally will be 312 electoral votes for Trump and 226 for Harris. And Trump @ 312 was by far the most common outcome in our simulations, occurring 6 percent of the time. In fact, Trump 312/Harris 226 is the huge spike you see in our electoral vote distribution chart:
[Interactive graph]
The difference between 20 percent (the share of times Trump won all 7 swing states) and 6 percent (his getting exactly 312 electoral votes) is because sometimes, Trump winning all the swing states was part of a complete landslide where he penetrated further into blue territory. Conditional on winning all 7 swing states, for instance, Trump had a 22 percent chance of also winning New Mexico, a 21 percent chance at Minnesota, 19 percent in New Hampshire, 16 percent in Maine, 11 percent in Nebraska’s 2nd Congressional District, and 10 percent in Virginia. Trump won more than 312 electoral votes in 16 percent of our simulations.
But on Tuesday, there weren’t any upsets in the other states. So not only did Trump win with exactly 312 electoral votes, he also won with the exact map that occurred most often in our simulations, counting all 50 states, the District of Columbia and the congressional districts in Nebraska and Maine.
I don't know of an intuitive test for whether a forecast of a non-repeating event was well-reasoned (see, also, the lively debate over the performance of prediction markets), but this is Silver's initial defense of his 50-50 forecast. I'm unconvinced - if the modal outcome of the model was the actual result of the election, does that vindicate its internal correlations, indict its confidence in its output, both, neither... ? But I don't think it's irreconcilable that the model's modal outcome being real vindicates its internal correlations AND that its certainty was limited by the quality of the available data, so this hasn't lowered my opinion of Silver, either.
Jump in the discussion.
No email address required.
Notes -
Is this the same Nate Silver, the mighty predictor? Or somebody impersonating him? https://x.com/RyanGirdusky/status/1855215191102750879/photo/1
More seriously, though, I just can't understand any meaningful way in which you can accurately predict Trump winning 312 electoral votes and then accurately predict his chances of winning the electoral college is a coin toss. These don't seem to be compatible in any sensible way. Maybe you can invent some statistical trick to make it sound good but on the plain common sense meaning it just makes zero sense.
Nate did nothing wrong. He is in the profession of polling, and must therefore believe that polls are directionally correct with some margin for error. He's then spent this entire life creating better priors, correlational models and ensembles to reduce that 'margin for error'.
A doctor doesn't question if germs exist. A mechanical engineer doesn't question Newton's laws. Similarly, Nate is incapable of questioning if polls contain any signal what-so-ever. Modern polling is in it's 2008 CDOs phase. You can't take 100 bad loans and roll them up into a AAA financial vehicle by citing diversification. Similarly, you can't ensemble broken polling data into any information of value.
Polls are useless for 2 reasons:
In a year with a wild card ex-president, incumbent president withdrawing, a VP who has never fought a competitive race, an assassination attempt, a fresh war in Israel, a lumbering war in Ukraine and a technically strong economy with terrible optics (lingering inflation from 2020-2021) ............ all your priors go into the dumps.
Even when polling does work (not often), it assumes a 'normal' year. In a "normal" setup, 3 things would have gone differently:
If these 3 hadn't happened, Trump would've still won. But, the Dem candidate could totally have flipped AZ, NV, Wisc & Mi. Still 2 short and PA was going Trump either way. However, in this world, Nate's predictions would have been a good proxy for the real results.
Alas, that never came to be.
Nate's 2023 victory dance is revealing. [1]
In writing this paragraph, Nate Silver fails to understand why the quote : "The plural of anecdote is not data" took off, and dooms his predictions for good.
[1] https://fivethirtyeight.com/features/what-the-fox-knows/
I'll leave you with my favorite stats quotes:
Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital -Aaron Levenstein
There are 3 kind of lies : Lies, Damn lies and Statistics - Mark Twain (maybe)
The plural of anecdote is not data - Not Ray Wolfinger
/images/17312889422478352.webp
I disagree. If he were a modest statistician that would just work his models the best way he can and produce the results, and let people interpret them as they will, without pretense, then he'd done nothing wrong. But as my link above suggests, he thinks his models reveal the deeper truth about the actual structure of the world, and his way of revealing it is superior to any other possible way - so much superior, that he is justified treating anyone who suggests the world may be different from what his models suggest with condescension and disdain similar to how a physicist would treat somebody who denies existence of Newton's laws. And the problem with it, of course, that this pretense of superiority is revealed, again and again, as false, and then complicated explanations are concocted why he technically has been correct all along and only by some weird fluke his opponents have been appearing to be correct. I think this is wrong. If you are in prediction business, and you predict wrong, you should at least eat the crow and be humbled. Otherwise you are in a scamming the gullible business.
More options
Context Copy link
More options
Context Copy link
You can easily imagine such forecasts
for example imagine this forecast
Trump wins every swing state but nothing else (30%) Trump wins Every swing state plus some extra (5%) trump wins split of swing states and wins election (15%)
Harris wins every swing state (25%) Harris wins split decision of swing states (20%) harris wins swing states + some extra (5%) (where swing is GA, NC, Mi, Wi, Penn, NV, AZ)
basically this forecast would be 50/50 and forecasts correlated polling error being a very strong effect.
More options
Context Copy link
More options
Context Copy link