[qqtt]

As someone who was super interested in the 538-style of election coverage in 2008, I've kind of fallen "out of love" so to speak with election models and forecasting in general. I'm not really convinced about what it adds to the conversation around elections. We can all look at various polls and get an assessment of who is generally ahead. Weighted polling aggregators and forecasting models just collect all these polls and spit out some data. It's easy to hand wave and think some new information is being revealed, but ultimately it is just a "garbage in garbage out" situation - you are entering polls as input, some hand waving is going on, and you get some forecast as a result.

I think part of my cynicism comes in the wake of the 2016 election, in which the forecast rightfully counted some scenarios in which either candidate could win, upon which conclusion of the model was basically "the result fits in with the forecast, because either candidate could have won according to the model" - in which case I personally concluded, if no matter what the result, we can always just say "the candidate who won could always have won given the forecast" - what are we really adding to the conversation here? We can simply look at polls and understand who is generally ahead, and not be any better or worse off.

[alisonatwork]

538 in its final form was not about predicting election outcomes, it was about the business and science of polling, and contrasting factual data with the way people perceived related issues. It was an interesting outlet that helped illuminate the data side of political science, and at its best also provided some insight into the disconnect between how the general public thinks about a topic versus what is actually happening.

Disney killing 538 is broadly a loss for political journalism in the US, imo, because most other American media is more interested in sensationalism and hyping imaginary culture war issues, i.e. exacerbating exactly the disconnect with reality that 538 was trying to combat with its more evidence-based reporting. From my perspective the only place still doing this kind of work outside of niche, single-topic outlets like SCOTUSblog is ProPublica, and even they don't tend to be as politics-centric as 538 was. So I definitely will miss the site, and the pod. I don't have the stomach for most other American media.

[wasabi991011]

Respectfully, I disagree on both points.

> We can all look at various polls and get an assessment of who is generally ahead.

I probably could, but there's a lot of polls to look through and I don't really want to spend the time. Much rather have someone else do it for me.

> if no matter what the result, we can always just say "the candidate who won could always have won given the forecast" - what are we really adding to the conversation here?

Isn't this hypothesis testing? If you have a weighted coin and a guess as to which side is heavier and by how much, you're going to need multiple flips to see if you are right. And it doesn't even really make sense to talk about how right/wrong you are about a single flip, only on the aggregate.

It's possible someone has already compiled FiveThirtyEight's results to get some aggregate accuracy, I haven't checked. If they have and he's wrong on average and that's what you are referencing, my apologies.

[JeremyNT]

The trick here is that if the election is close enough that you'd actually want/need multiple pollsters aggregated, the aggregators will indicate high uncertainty. If it's enough of a blowout for the aggregators to indicate low uncertainty, then the individual polls are going to be showing a large gap.

An aggregator saying "foo has a 65% chance of winning" may seem like it's providing more information than a single historically reliable poll (say Reuters/Ipsos) stating "foo is up by 2 points but there's a 3 point margin of error" - but isn't it just an illusion? High quality pollsters very seldom deviate very much.

And even if you grant that the aggregator is closer to being "right" than any single pollster, is that difference actually meaningful enough to impact any real world behaviors? Would you do anything differently with a 50% chance of victory versus a 70% chance?

I've honestly come to think of them as entertainment, with no real value.

[tqi]

I know the folks at 538 meant well, but I think the ultimate impact of their work was to accelerate the politics as entertainment, team sports-ification of elections, to our nations detriment.

[sdwr]

I kinda get your point - statistics suck the air out of the room. If regular people are talking about swing state poll margins of error instead of the actual issues, something's gone wrong.

538 democrasized the numbers that were the domain of political whizzes. I don't know if that's a good thing.

[maigret]

Well swing states are the issues, aren’t they? I’m happy to live in a place where it isn’t a small fraction of individuals deciding the fate of the country.

[wslack]

I think aggregators are useful to tell us what actually moved voters in a campaign - the swings are visible even if baseline error is not.

[bee_rider]

Eh, this seems to just promote armchair quarterbacking. What moved voters is an issue for the campaigns to track. We should listen to what the campaigns say about what we care about.

[bnralt]

Like the Rationalist's "Bayesian priors," the election models were a remnant of the "big data" hype from a decade and a half ago. This article is a decent overview for anyone who forgot about it[1]. Like with many hype cycles, there was something actually important underneath the surface (useful statistical modeling), but then people with a poor understanding of the limitations ran wild thinking it could do things far beyond its capabilities (in this case, the degree to which one could use statistics to predict the future).

Industry gave up on the more extreme claims fairly quickly because it wasn't able to produce. But it lingered on in other places where there was less direct feedback or it was telling people what they wanted it to hear.

To add to this, it became obvious that many of the leaders in this "field" were people who believed they had an expertise that was far beyond their actual capabilities. Nate Silver ended up accusing much of the polling industry of fraud recently, because he wasn't able to do basic statistical math[2].

[1] https://slate.com/technology/2017/10/what-happened-to-big-da... [2] https://x.com/JustinWolfers/status/1853302476406993315

[Gimpei]

I disagree. Before 538 people were still offering lots of election predictions and it was much much worse, because it was based entirely on hunches and vibes. Silvers rates the pollsters and provides confidence intervals far better than a simple average of polls does. I’d much rather read his forecasts than any number of bloviating opeds.

[mgfist]

But that's because elections these days are incredibly close. It's like being upset that the best statistical answer to "who will win a coin toss" is "well it's 50/50".

[genewitch]

Are you sure about that?

89% of counties turned red. Looking at 1.5% total difference is missing the forest for the trees. Trump got all swing states.

To quote: "this is absolutely a mandate."

https://youtu.be/zG3n2IeaTPA

edit: facts cannot be insolent. The youtube link is a guest on the view named Stephen something, from this week, saying the words that i typed into the comment box. Not clicking it is doing yourself a disservice; as it is "source cited."

Don't get mad at me for relaying this information.

[roenxi]

Sure. And what was the probability of something like that happening? About a coin flip.

It is an election using First-Past-the-Post counting. A 50:50 probability doesn't mean the final count will be close, it (usually, anyway) means there is evens odds which side is about to win decisively. The county results are expected to correlate, as are the swing states in all likelihood.

[genewitch]

the probability of 89% of blue counties (in 2020) switching to red counties is a 50-50 chance? can you show your work?

[roenxi]

https://projects.fivethirtyeight.com/2024-election-forecast/

You'll note that the distribution mode for a Trump win was 312 electoral votes; which is what he got. They do a good forecast - it suggests there was a 50-50 chance Trump wins, and if he wins the most likely outcome is ... exactly what happened.

If it was a Harris win the best bet is she would have gotten 319 electoral votes with high correlation in the swing states too.

[genewitch]

you still didn't watch the video i linked^, and you did not answer my actual question: 89% of counties turning red compared to the 2020 election is a 50-50 chance? Can you show your work?

^ The person in the video explains why thinking it is 50-50 chance is detrimental to the liberal cause.

[aaronbrethorst]

Counties don't vote and Trump's popular vote "mandate" is smaller than Clinton's popular vote win in 2016.

Here's a fun stat: literally 40 states have a population that is less than the population of Los Angeles county alone. Why doesn't Los Angeles itself have 80 senators?

https://en.wikipedia.org/wiki/2016_United_States_presidentia...

https://en.wikipedia.org/wiki/2024_United_States_presidentia...

Setting aside 2000 and 2016 (EC winners lost the popular vote), you'd have to go back to 1968 and Nixon's squeaker of an election over Humphrey to find a closer popular vote: https://en.wikipedia.org/wiki/List_of_United_States_presiden...

[JumpCrisscross]

> why doesn't Los Angeles itself have 80 senators?

The electoral college internally balances the power of a population against the difficulty of holding land. Look at New York State, where NYC mostly holds court. If it were a country you’d see rebellion. Because while the city outnumbers the country, it’s culturally more similar to itself than the country, and that in practice leaves lots of people disenfranchised.

(Personally, I think the President should be popularly elected. But the Senate should continue resembling our geography.)

[int_19h]

Even if you buy into the whole notion of "representing geography", the problem is that state boundaries ceased to be representative of any meaningful kind of political distinction a long time ago, as evidenced by the massive red/blue splits in many states. It's not just NY - you can see the same thing in other large states, e.g. in WA where the split is geographical within state boundaries - west of the Cascades is very blue, east is very red. Nor is it unique to blue states - TX has the same exact issue with blue counties having a lot of population that is effectively not represented at all.

Unless and until this is fixed, there's no meaningful "geographic representation" in the Senate, so it's strictly a negative.

[skissane]

> Here's a fun stat: literally 40 states have a population that is less than the population of Los Angeles county alone. Why doesn't Los Angeles itself have 80 senators?

Because in a federal system it is often considered important to provide the less populous states with some protection against the more populous states always getting their way. The US Constitution does this by balancing representation based on population (the House) with equal representation of each state (the Senate).

The US is not the only federal constitution to do this - the Australian constitution has the same design (indeed, copied off the American model), except having 6 states instead of 50, Australia went with 12 senators per a state instead of only 2 - hence Tasmania (population 571,200) gets 12 senators, and so does New South Wales (population 8.153 million).

Things don’t have to be this way - instead of a federation one could have a unitary system. But in the case of both countries, protecting the power of the smaller states was considered important at the time of the constitution’s drafting - and the smaller states likely would not have agreed to it otherwise

Where it is different, is Australia doesn’t have the same “red state” vs “blue state” dynamic the US does. In Australia, while some states lean more one way than the other, they essentially all are “swing states”

[aaronbrethorst]

"in a federal system it is often considered important"

by whom?

Definitely not James Madison who only grudgingly accepted this framework in Federalist No 62: https://avalon.law.yale.edu/18th_century/fed62.asp

[skissane]

> by whom?

Well, both by the majority of the drafters of the US Constitution, and the majority of the drafters of the Australian Constitution.

And the authors of the German Constitution – the German upper house, the Bundesrat, represents the German states (Länder), and although (unlike the Australian and US Senates) it does give more populous states a greater number of seats, the number of seats is still out of proportion to population: in Bremen there are 223,830 people per a seat, compared to 2,977,586 people per a seat in North Rhine-Westphalia.

And the authors of the Swiss Constitution – the Swiss upper house (Council of States) gives two seats each to twenty of the country's cantons, and one seat each to the other six (which six are traditionally referred to as "half-cantons")

And I'm sure I could dig up more examples – globally, the majority of federations have an upper house which provides, either equal representation to each state/province, or if not equal, then at least representation that deviates significantly from proportionality to population.

> Definitely not James Madison who only grudgingly accepted this framework in Federalist No 62

I'm not sure if Madison should be interpreted as "only grudgingly" accepting this framework – but even if that's true (Madison was very much an advocate of centralized power and supporter the interests of the big states over that of the smaller states), many of the other delegates at the 1787 Constitutional Convention viewed it more positively, even as a necessity – the majority of delegates agreed to the Constitution containing this provision, and it is unlikely such a majority would have if it had been ommitted.

[genewitch]

Out of curiosity, do you recall seeing a "county map" of the blue votes in the 2020 election?

[danorama]

I don’t personally recall, but it’s kind of a silly idea to break down presidential votes by county anyway. At least counting states matters legally in elections. And counting people matters as a way to judge desires and sentiments.

Counting by counties combines the unrepresentativeness of the Electoral College with the legal irrelevance of the popular vote.

[kemayo]

Why would we have? It's pointless. As the GP said, land doesn't vote.

The more interesting map is the one where you weight the display by population, since it gives a much more accurate representation of how the USA is divided: https://engaging-data.com/pages/scripts/d3Electoral/countyel...

The tool on that website is actually pretty informative to play with if you want to quickly see how adjusting the map style can change the perception. https://engaging-data.com/county-electoral-map-land-vs-popul...

[madcaptenor]

Not sure why I can't reply to the sibling comment...

anyway, the reason for the east/west divide in population density is that the west is dry. See maps at climate.gov (https://www.climate.gov/news-features/featured-images/new-ma...). By convention people refer to this line as being at the "100th meridian" but the 98th may be more accurate.

[verzali]

Bit of a tangent, but I've always wondered why the United States has a very clear dividing line between the populated east and relatively empty west. It's really visible in the first map you shared. Is there a historical reason behind that line? Or some geographical boundary?

[seanmcdirmid]

If you include Alaska, which is a huge state (I hear bigger than Texas), land starts voting very blue due to Alaskan native influence. They don’t have counties though, just boroughs (Alaska still goes red even if it’s land goes blue due to population distribution).

[genewitch]

Louisiana doesn't have counties, either. I live in Louisiana, and when i said "counties" i included the parishes in my state, and any other geographically distinct entity of the same legal merit.

there are 3300 counties in the US. People live in all of those counties.

[aaronbrethorst]

Did you know that there are about 2,160 counties across these 40 states that collectively have a smaller population than Los Angeles county? Who cares about a "county map"?

[genewitch]

people who don't live in Los Angeles might be curious why their neighbors suddenly switched to voting red.

why is everyone assuming that i'm not intelligent enough to know that "land doesn't vote"? The county voting maps are a proxy.

[nobody9999]

>https://youtu.be/zG3n2IeaTPA

>note: not clicking that because you disagree with me is really doing yourself a disservice.

I won't click the link because I'm here for discussion with the users of this site, not some random video you thought was interesting/useful.

Why do you find it interesting/useful? What arguments are made? How do those arguments comport with your beliefs?

Let's have a discussion. So tell me what you think, don't link to some rando on youtube.

[bee_rider]

Also it is a video. Asking literate people to go watch a video is disrespectful of their time.

[genewitch]

It is a video of a guest on the view saying the words that i typed. It was "citing a source."

[bee_rider]

The View is just a chit-chat show anyway, so a source that proves that somebody said something on it doesn’t prove anything compelling.

Anyway, my computer can’t play YouTube (ad blocking issues).

The underlying thing that is annoying people, I think, is that this is (as far as I can tell based on what you’ve written) a factual mathematical issue. That sort of thing is easily conveyed via text and equations. Why would anybody want to watch a video of somebody explaining math?

[genewitch]

lol TIL "The View" is "some rando on youtube"

And to answer your question, the video is a guest on the view who i don't know, saying the exact numbers i quoted. I was quoting the person on the view, which aired yesterday or something.

what's interesting, is instead of discussing what i said prior to the link, you chose to instead try and make me feel bad for linking a video.

Keep it classy, HN.

[leourbina]

Anyone can say anything on YouTube. I think you’re confusing confirmation bias with “citing your sources”.

[genewitch]

you didn't watch it either, huh? Nice.

also what bias? I didn't make the stats up. Statements of fact cannot be insolent. If you disagree with the facts, then let's talk about that, and look for more data. However, what has happened in this thread is thinking i have an agenda, other than i think people need to hear what they were talking about - in that clip

The mentality that "it was real close to 50-50 and trump didn't get a majority of the vote and therefore we can keep on doing what we're doing" is what the video clip was talking to. It didn't work, and the tactics need to change if anyone wants to see change.

I don't care that people want to argue with me personally. But it's not doing themselves any favors, as it pertains to getting people elected they want to see in office.

[nobody9999]

>what's interesting, is instead of discussing what i said prior to the link, you chose to instead try and make me feel bad for linking a video.

Others had already rebutted your assertion at least as well as I could, so I didn't feel the need to repeat what had already been offered.

However, given your initial rationale:

>>note: not clicking that because you disagree with me is really doing yourself a disservice.

was a poorly constructed straw man. Which I noted. It wasn't that I was rejecting you, I was clarifying that I (and likely many others) come to HN to discuss matters of interest to us.

If I wanted to watch videos, I'd go to youtube and the like. I came to HN instead.

I'd point out that you didn't make clear that you were "citing your sources" with the video link.

Now that I know the source ("Stephen Somebody or other" who managed to get himself booked on some low-information blab fest to make his important pronouncement), my initial response, "[s]o tell me what you think, don't link to some rando on youtube," was spot on.

All that said, linking to video sources is absolutely reasonable. In fact, I've referenced stuff from videos several times.

But each time, I made sure to explain the context of the video, the text of the quote and, most importantly, who was being quoted.

As I did here:

https://news.ycombinator.com/item?id=30762760

[genewitch]

Okay. have it your way.

The thought that the vote was "real close to 50-50" and "trump didn't get a majority of the vote" and "therefore we just need to do what we're doing and it'll work out OK in 2028 and the midterms" is what the video clip i linked was talking to.

Specifically, nearly every reply to my comment, other than yours, argued that "the number of counties that switched" is irrelevant, as if that happened by accident, as if your neighbors apparently changing from blue to red for the 2024 election isn't a bellwether of something else. Trump still got a plurality of votes. Asking "why" is something that needs to be done.

Nearly every comment assumed something about me, because i quoted a statistic. I knew, because i have been on internet forums for over a quarter century, that no matter how i phrased my comment, i was going to get downvoted and argued with.

[kemayo]

I don't think someone can claim a "mandate" based on not even having won 50% of the vote. You need to at least be able to say that over half of the voters wanted you to win in order to use language like that.

That 1.5% is the only thing that mattered for "mandate" purposes. Trump won, but on a razor's edge, and that should give him some pause. (It won't.)

[mgfist]

Well, the great thing is we have betting markets for this now. Given your confidence, I'm assuming you made a lot of money betting on Trump?

[nozzlegear]

I largely agree with your points. Election modelers and forecasters really don't add much to the conversation after 2016, despite their attempts and even purported success at correcting their models and mistakes. The only election forecasting model that I take seriously these days is my own vibes based forecasting.

I have enjoyed the meta-drama around forecasting and modeling that pops up every election season though. It's hard to beat "[Nate Silver] doesn't have the faintest idea how to turn the keys," or "I ran 80,000 simulations."

¹ https://knowyourmeme.com/memes/he-doesnt-have-the-faintest-i...

² https://knowyourmeme.com/memes/i-ran-80000-simulations

[JumpCrisscross]

Would note that Silver owned the election model 538 ran on. When he left, he took it with him. The recent election forecasts 538 put out were not Nate Silver’s. (In my opinion, his were more accurate. More importantly, his commentary was more informative on the model’s shortcomings and insights.)

[finnthehuman]

>if no matter what the result, we can always just say "the candidate who won could always have won given the forecast" - what are we really adding to the conversation here?

When the "upset" candidate wins, it will have been statistically likely. Yes. What they're adding are error bars in the collective consciousness.

Ironically enough, the hype pop of 538 was actually driven by people misinterpreting the stats and using them to feel vindicated in their support of Obama. The comedown was finding out that 25% is still 1-in-4 odds.

538 should be killing the horse race coverage by doing the most sober version of it. But horse race is big business. Therefore: boo 538, booooooo, they're harshing the vibe of my favorite reality show.

[smt88]

The purpose of statistical coverage of elections is to tell campaigns what their strengths and weaknesses are and to entertain the rest of us.

Since we don't run campaigns, the only value is entertainment. But that's not a negligible value for a lot of people.

[ks2048]

I agree it's mainly entertainment, but it also affects a lot of people "involved" in campaigns who aren't "running" them - mainly people donating their money or time.

[CalRobert]

I find it useful in the general way that having corroborating sources for election data is useful. If the election is reasonably close to models, exit polls, etc. then it is more likely to have been run fairly.

[mikepurvis]

As a resident of Kitchener ON, I'm privileged to live in one of just two Canadian ridings represented by the Green Party— I worked on Mike Morrice's first campaign in 2019, and I remember being frustrated trying to talk to voters on doorsteps and having to explain over and over that the purported "polls" on 338 (Canadian knockoff 538) showing us at 3-4% support in the riding were based on projecting forward previous results and adjusting slightly for national polling trends.

Sure enough, we ended up coming in at a whopping 26% in 2019, and in 2021 won the seat with 33%. Certainly the win was in part because the incumbent was embroiled in a last minute scandal, but I truly believe the polling aggregators have a huge suppressing effect on breakout candidates— without that effect it's possible we could have taken the earlier election too.

Now that seat is "safely" Green, it's been twice affirmed with huge wins for a separate Green candidate at the provincial level:

https://en.wikipedia.org/wiki/Kitchener_Centre_(federal_elec...

https://en.wikipedia.org/wiki/Kitchener_Centre_(provincial_e...

I expect this year's federal election will deliver another 40-50% result for Morrice, as he's very popular locally, but there's 338 again showing a big upswing for the Liberals in Kitchener Centre, when almost certainly there is no such thing, it's all just hallucinated from national polling:

https://338canada.com/35048e.htm

[Uehreka]

I find the modeling super useful, many conventional media outlets still don’t properly communicate probabilities to their audience. For instance, I vividly remember the following exchange between Nate and some news anchor in during one of the 2016 conventions:

Anchor: So Nate, you say Trump has a 25% chance of winning, can you tell us exactly what that means?

Nate: Sure. So imagine I flipped two quarters, and they both came up heads. In that scenario, Trump wins.

Anchor: (shocked) Wait but that’s… that’s a thing that actually happens! You’re saying Trump has an actual chance of winning?

Nate: Well I’d rather be Hillary than Trump right now, but yes, people shouldn’t be that surprised if Trump wins, his chances aren’t insignificant.

I remember people in October still saying that Nate had to be wrong, that there was just no way Trump could win. There was even a growing market for what I would now call “cope forecasts” that “unskewed” the results to show that, really, Hillary had a 99% chance of winning, just like you knew she did (all of these people looked extremely foolish after the election was over).

I also feel like good models provide valuable pushback against media narratives that try to characterize the “closeness” of a race. In 2016, people wanted to hear that Trump had no chance of winning, but Nate/538 correctly pushed back that the race was actually pretty close and both candidates had a good chance. And he did the opposite in 2012: Pundits wanted to cast Obama and Romney as being neck-and-neck (which makes for a more exciting story) and Nate had the stats to push back that actually the race was not very close at all. If Romney had won in 2012, Nate would’ve had to eat crow, but Romney didn’t win.

Nate and 538 also do senate races, which are super valuable if you’re figuring out which candidates to donate money to. Often there are Democratic candidates in totally doomed races against Republicans I really don’t like, and the data helps me look at those situations and go “yeah I hate Lindsey Graham, but his challenger has no chance, I’m going to donate to the milquetoast Nevada senator whose race is on a knife’s edge instead”.

I could probably just look up polls, but the way Nate/538 process the polls into results with error bars and probabilities makes it a lot easier to reason about.

[bnralt]

I see this argument a lot, but it's contradictory. You're simultaneously arguing that people don't understand statistics because they're treating a 25% chance as no chance to win, but then you're doing the same by saying that the other predictions, in the 15% to 2% range[1] are "cope forecasts" that people who followed them "looked extremely foolish" (the only major 99% forecast was PEC, but Wang said that's because the model broke down and the actual forecast was around 5% [1]).

25%, 15%, 5%, even 2% chances happen with a decent amount of frequency. I don't understand how people can say that people don't understand probability because they think a 25% chance won't happen, but then turn around and treat a 15% chance the very same way.

[1] https://www.nytimes.com/interactive/2016/upshot/presidential... [2] https://web.archive.org/web/20171120175008/https://election....

[MostlyStable]

I agree with basically this whole comment, but the sad/ironic/whatever part of it is though even though Nate Silver was doing the thing you are describing ("actually, Trump has a real shot of winning"), afterwards he got constantly dinged (either by people incorrectly conflating him with the 99% models, or by people who just didn't actual listen to him) for "getting it wrong" because he "only" gave trump a 30ish % chance of winning.

I still come across it every once in a while and it probably has the highest ratio of level of infuriating-ness to low-value of the stakes of the opinion of just about any political opinion I can think of.

[fitsumbelay]

agreed on all points

[tbrownaw]

But how else will you know the correct emotional tone to use when complaining / gloating about the upcoming election with friends? Or know how seriously to take various prepper activities like stocking up on Twinkies in case the wrong team wins?