The Electoral College is growing larger
In 2016, the Electoral College favored the GOP by about 2.9%, the difference between Hillary’s popular vote margin and Trump’s margin in the tipping point state (Wisconsin.) The conventional wisdom was that the EC bias would be about 2% to 3% in this election:
If you insist on a point estimate, it looks like Nate Silver is at around 2.6% or 2.7%. But in October, I argued that the EC bias was growing and would be even larger this year, say about 3.5%:
It’s probably pointless for me to predict the presidential election, given that I’ve lost touch with American politics. (I got them all correct from 1968-2012, before missing badly in 2016.) Instead I’ll just predict a very strong pro-Trump EC bias, even stronger than in 2016 (when it was about 2.9%.) Just to make things interesting, let’s say an EC bias of 3.5%. I base that prediction on the assumption that the Dems continue to gain ground in states that they would likely lose in a close election, such as Texas and Georgia. (They might win Georgia, but they’d lose it in a close election.)
[Not bad! Biden will probably eke out a narrow win in Georgia, but would have lost it in a closer election.]
We don’t yet know for sure, but it seems highly likely that Wisconsin will once again end up being the tipping point state. Silver estimates the national popular vote margin will be about 4.3% when all votes are counted, and that implies an EC bias of 3.7% (given that Biden won Wisconsin by 0.6%.) That’s huge.
The shy voter bias was even larger than I anticipated. The 538 website had a pre-election margin of 8.5% in the national polls, and a “model” forecast of 7.8% (I believe incorporating state polls as well.) Even using the conservative 7.8%, that’s a 3.5% miss in the national polls, much worse than last time.
Add the EC bias and the shy Trump voter bias, and Biden needed 7.2% of that 7.8% poll lead in order to win this election. Back in September when Biden led by 7%, the race was probably even, which is why the betting markets had it even at the time. My Trump victory prediction in September was obviously wrong, but not as crazy as some commenters might have assumed.
Biden really did need at least a 7.2% lead in the polls. And even this may understate the shy Trump voter bias, because many votes were cast in mid-October, a period during which Biden’s poll lead often exceeded 10%. The average vote was probably cast at a point in time when the poll lead was roughly 9%!
Trump’s almost unbelievably inept campaign might actually have hurt him.
And then there’s Scott Adams, who should have quit while he was ahead.
PS. American elections are annoying in many different ways. Here are just a few:
1. Vote counting slows down each day, for no obvious reasons. In places like Pennsylvania and Georgia, the vote count today was much slower than yesterday. Why? Just get it over with. The final vote count probably won’t be known for weeks, which is a disgrace. Florida is the nation’s third most populous state, and had almost all votes counted on election night.
But then why be surprised? We are a banana republic.
Update: Razib Khan has a nice graphic.
2. Networks call states for no apparent reason. Fox News and AP called Arizona days ago, despite no clear evidence that Biden had won. It was just a guess. In contrast, it’s obvious that Biden has won Nevada and Pennsylvania, and yet the networks have not called those states. Networks should establish a consistent procedure and stick with it. I could do better than the clowns they hire.
If you don’t know, then don’t call it. How hard is that?
Update: Biden will apparently be the first challenger to get 51% of the vote since FDR in 1932. And in a year when most voters gave the incumbent good marks on the economy.
Tags:
5. November 2020 at 22:41
Maybe the polls are not wrong.
The difference between polls and vote counts could also be explained by voter suppression. The increasing electoral college margin could also be explained by extra efforts at voter suppression in key states.
6. November 2020 at 00:42
The slow down in counting might just be for pruely technical reasons. Eg they might start with the easiest to count votes?
And obviously they are counting the votes for different areas of the state in parallel. And some areas will finish before others.
Areas that are already finished won’t contribute to the votes counter per hour metric anymore.
Of course, those two factors are fairly universal and don’t explain any differences between states.
6. November 2020 at 00:50
I will say this: Even if the Democrats stuffed or unstuffed ballot boxes in several key states (entirely possible, given human nature and copious historical precedents by both parties), the results of the national popular vote are for Biden.
The Electoral College has become black comedy, installing losers, or nearly so, into the White House in each cycle.
Maybe it is rough justice that Biden will go to the White House. He still belongs on Gilligan’s Island, but hey, he won the vote.
The Biden coalition is creepy, but we have seen worse in the US.
6. November 2020 at 05:10
“Vote counting slows down each day, for no obvious reasons. In places like Pennsylvania and Georgia, the vote count today was much slower than yesterday. Why? Just get it over with. The final vote count probably won’t be known for weeks, which is a disgrace. Florida is the nation’s third most populous state, and had almost all votes counted on election night.”
It also gives conspiracy theorists ammo. I saw a tweet that said “counting ballots takes a lot longer when you have to personally fill out the ballots as you go”
6. November 2020 at 05:19
Scott Adams is a sad case. I used to like Dilbert but I guess he’s just a one trick pony. I can’t even look at Dilbert anymore. The problem is not that he’s a Trump supporter, people may have some sophisticated reasons for that (not that I know any). But Scott Adams is just deluding himself with mediocre metaphysical nonsense and doesn’t even have the guts to say what’s probably the truth, that he (as a cartoonist) probably just admires Trump’s verbal gift with insults. Instead he’s just smugly pretending that there’s some depth to the man.
6. November 2020 at 05:25
If anything, we are probably underestimating the size of the bias. Big GOP states are disproportionately competitive compared to big Democratic states, and GOP therefore already needs to turn out more of its voters than do the Democrats.
GOP is already having to turning out its folks in FL, TX, OH, GA, PA, NC, MI, etc. Democrats aren’t having to do so in NY, CA, and IL.
Of course, the counterfactual would also need to think about the potential increased turnout of GOP voters in clear majority states like NY, CA, IL (Or Democratic voters in, say, Oklahoma, although they are fewer)
6. November 2020 at 05:37
OT: unemployment rate fell to 6.9%. Looks like Lars Christenson’s <6% November unemployment prediction might end up being right.
6. November 2020 at 06:25
It slows down for very obvious reasons. They need to know how many votes their candidate needs. Two USPS workers were caught on video admitting to backdating votes and filling out empty ballots. 22000 dead people found registered to vote for Biden in PA. 6000 non residents found voting in Nevada. And that is just the tip of the iceberg. As they dig deeper, they will reveal more and more fraud.
Trump will win in the end.
Biden just destroyed the democratic party forever with these shenanigan’s.
6. November 2020 at 06:28
Sumner is spot on about the timing of predictions. Annie Lowrey has a nice summary of why Trump’s re-election prospects spiked in October, a spike few (nobody?) predicted. To think that if the election had been delayed a few weeks Trump may have cruised to re-election. What does that say about the American voter? One would guess that Trump would have tried to delay the election if only he had known what we know now.
6. November 2020 at 06:56
They shouldn’t be calling any states until these lawsuits have concluded.
Nevada has now 6000 identified voters for Biden who don’t live in the state. ‘
Whether you are democrat or republican, that should really concern you. If it doesn’t concern you, then you have clearly been brainwashed by partisan media outlets.
When you have 128000 ballots showing up at 4am in the morning, while ballot counting has stopped, and all 128000 votes are for Biden that should be concerning to ALL Americans.
For those that are math illiterate, the probability a box of 128000 ballots show up at 4am and all happen to vote for Joe Biden and no other candidate (there were no votes for down ballot candidates) is zero. It cannot be duplicated in reality. You could tell 128000 people to vote for Biden, and the statistical probability of all 128000 succeeding is close to zero. I.e., someone will make a mistake.
Lots of questions the democrats will need to answer in court. With videos emerging of usps workers admitting to fraud, this doesn’t look good for the democratic party. My money is still on Trump winning this.
6. November 2020 at 07:34
Janice,
If you are serious about “your money”, lets bet! If Trump wins I give $100 to a charity of your choice. If Biden is sworn in in January, you give $100 to doctors without borders.
Deal?
6. November 2020 at 08:01
Democrats are inciting violence now.
The law in every state allows poll watchers, and for some reason PA is refusing to allow any poll watchers to stand within 6 feet of the table, despite a court order. This will backfire. And if it doesn’t, you can bet that we headed to civil war. The notion that a state in this Union can secretly count ballots, and defy court orders, is a significant problem.
I hear militias are already planning to attack if Supreme Court fails to hold states accountable for violating laws.
Dangerous times!
6. November 2020 at 08:29
Speaking of Wisconsin.
At 10 am this morning the USA Today site showed Biden had 1,630,568 vs Trump 1,610,029.
But the aggregate Democratic (1,567,733) and Republican (1,660,284) house candidate vote was significantly in favor of the Republicans.
It seems like people are rejecting Trump but not necessarily Trumpism (at least in Wisconsin).
6. November 2020 at 08:32
Couple notes:
In Pennsylvania, the Democrats wanted to move (finally) to a system like in FLA and count the mail in ballots in advance. They should have proposed that years ago. The Republicans blocked that. I hope they revisit that now.
Your observation about missed Trump support by pollster looks accurate. But I highly doubt it’s due to shyness. The name for that phenomenon should be changed (by everyone – I know that you are just using the common phrase).
Wow! That was a great call about Georgia.
Are you sure that Wisconsin was the tipping point? I want to wait and recheck that once the rest of the states were called.
Banana Republic phrase is so apt. And sad. For me, I see this as just a 2-4 year hiatus. Trump supporters can take some comfort here that they will have plenty of chances in the future.
Civil War? No way. I won’t bet on this because I wouldn’t give me contact info to anyone who would bet against me. But I’d give 10 to 1 odds against a thousand people dieing in political violence between now and January 31st. Like I said though, I won’t back that up with a real bet (unless one of my close friends sees this and calls me 🙂
6. November 2020 at 08:42
Olivier, You are mixing up WHETHER the polls are wrong with WHY the polls are wrong. They’re wrong.
Garrett, Wow, that unemployment figure is impressive.
You conspiracy nuts, LOL.
6. November 2020 at 08:47
I am very surprised American people cannot see the corruption. Maybe too much propaganda by MSM outlet.
I even hear MSM talk about removing Trump, but military is 90% conservative. And the chief of staff is also conservative. That is talk of coup. Its talk of totalitarian ownshersip of govt. Courts should decide. Not MSM media. And military will never remove him. They love him. Because he loves the country. They are so out of touch with American people behind that desk.
When trump was winning yuan collapsed. Now that biden has retaken lead yuan is soaring. That is proof ccp owns biden. Investors know it.
6. November 2020 at 09:01
Same thing seems to be happening in Georgia.
Trump 2,448,565
Republican house candidates 2,477,352
Biden 2,450,512
Democratic house candidates 2,370,633
It appears Trump is losing because he’s Trump.
6. November 2020 at 09:24
the only part of Trump campaign that was excellent was second debate—-his rallies were opportunities to shine—he did—but he also unshined when it was unnecessary. I am surprised it is this close since I do give the + on fraud to Dems. Ga., as you now know is doing recount.
This should not have been not a Biden landslide like Nate implied—but a solid Trump win. But DJT was driving me crazy and I support him. So on the one hand I am surpised it is this close, even though it should have been a layup.
Still, on the conspiracy front, The 6 states which are still technically in play have a total of a 10k vote difference in Biden’s favor—so says RCP at this time.
the odds that these close states would all go to Biden–which is the current belief (because PA is still “counting”) with this total is like 6 heads in a row coin flip all for Biden–so about 1 out of 50–which happens 1 out of 50 times :-). P-value of .02. It only means it was unlikley—very unlikely—but not impossible.
The fact that there was high serial correlation in the counting after about 80% of the vote in all 6 states—but not the other 44 states–is also strange —and if you narrow it down to Mich, Wis, PA, and Ga—-it is so high that it decreases the p-value at least by half. If late count were “always” going to favor Biden, why in Democrat cities in THESE states did it happen to go that way but not elsewhere.?
Not proof, maybe not even evidence—but two of the states are already having recounts
6. November 2020 at 09:36
Jim, I agree.
Michael, Just stop with the “data analysis”. These margins will get much bigger. Biden is not nearly as “lucky” as you imply.
6. November 2020 at 09:39
Plus—if you, Scott Sumner, knew the Congressional counts before hand, what would you think the odds would be that Trump would trail in contested states? Most people thought–or so everyone says now—the lower he went the lower they would go—-but my prior was he would have won–what was yours under those conditions/
Conspiracy is a high probability event—successful conspiracy is a low probability event (think FBI, Russian Collusion)–same in voting. But the probability of success rises as the degree of conspiracy required declines. I am glad we are having recounts–as I was in 2000. Then when Biden wins—we can all (people like me) shut up.
6. November 2020 at 09:48
Bill, I’m pretty sure that in PA, the GOP passed a bill to count the votes early and to authorize election officials to reach out to voters whose absentee ballots didn’t meet requirements and offer them a chance to cure. The governor threatened to veto it, and they were never able to reach agreement.
The bill included some provisions the governor didn’t like, such as a limitation on how far before the election voters could request ballots and a requirement that absentee dropboxes be limited to boards of election to provide for security.
You can argue about whether the GOP security measures were reasonable and whether the governor was reasonable to turn them down, but I think it’s equally fair to say that the GOP proposed to count the votes early and the Dems blocked it.
Agreed that I hope they work it out going forward.
6. November 2020 at 09:54
This is a pretty good article on why PA is so slow to count.
https://www.inquirer.com/politics/election/spl/pennsylvania-election-2020-counting-results-delays-mail-ballots-20201104.html
6. November 2020 at 09:56
The urban/rural divide is a real issue. And maybe a reason to be in favor of the electoral college.
Urban areas have international platform businesses with high margins (tech, hollywood, finance). Rural areas are often in more commoditized businesses (farming, small industry). Wealth inequality matters more to cities as you have billionaires, high paid middle managers, and the service workers serving them.
Policies that might sound great in blue cities and maybe on a utilitarian basis be beneficial could be disastrous to rural America. When a good manufacturing job in the rust belt pays $20 an hour and the business only has 10% margins there’s no money to pay out a $15 minimum wage. And less need due to lower costs of living.
6. November 2020 at 10:28
Scott–that is exactly my point—–right now it is 50 50. Assuming Biden will win 60% of the remainder, in a coin flip world the odds of that happening is literally 1 out of many millions. That is what will happen. So it is obviously not a coin flip world. The reason it is not a coin flip world is the remaining counts are known to be Biden areas. This makes perfect sense.
What does not make perfect sense is why in these states are the known Biden votes counted last? Not all of them of course–many were counted early as in before polls closed—-just enough of them were not–many mail in of course. —the amount needed–by coincidence.
Except we know some of these cities (Detroit and Philly) are absurd. its not crazy Scott—
And using your logic on PA versus the rest of the country—it means he underperformed by even more in the areas where they won congressional seats—-To say this was surely unexpected is true—a
But it never should have come down to this–he blew it in plain site
6. November 2020 at 10:30
J Mann,
Thanks. I will check that out.
6. November 2020 at 10:39
Michael, You said:
“The reason it is not a coin flip world is the remaining counts are known to be Biden areas.”
No that’s not the reason. Why don’t you become informed before wasting all our time with your comments? The late counted votes are mail-in votes, which tend to be Dem votes from all areas.
Trump told his supporters not to vote by mail—and now he complains!
6. November 2020 at 12:45
“Trump told his supporters not to vote by mail—and now he complains!”
That was actually his strategy. Get most of his supporters to vote in person so that their votes would be counted first. Then, after their votes get counted, insist that the counting stop before the mail-in votes could be counted.
He was correctly warning his supporters not to trust mail-in voting. He just didn’t explain that the reason why was because his own efforts would be directed towards invalidating those ballots.
6. November 2020 at 13:09
The big excuse in 2016 was Hillary’s emails. Comey’s October surprise No one can argue that something dramatic happened over last week. Biden’s lead was both large and stable….thru 2 debates, Trump’s taxes, pandemic, and Hunter’s laptop. It wasn’t one fluke poll. North Carolina Senate polls….no movement despite Cunnngham sexting “scandal”…no one forecast Biden/Collins in Maine yet everyone finds split ticket voters AFTER the fact. My theory is that pollsters are either lazy or fraudulent….40,000 election simulations updated every 45 minutes ! (That comes to roughly 800 hours of my life I’ll never get back.). Thanks Nate
6. November 2020 at 13:15
Scott—that is not totally true—PA (Phil, Allegheny) Arizona (Maricopa, Pinal) and Nevada (Clark and Washo are the two big ones) clearly are county driven —don’t know about NC. . Georgia and Wisconsin have already agreed to recounts. Almost all counties in Nevada are counting–mostly Trump and mostly small. Clark is Biden and obviously big. So I don’t know where the mail-ins are from—but presumably from the counties with less than 100% counted. So they may be mail in but they are also Democrat.—i.e NOT ALL AREAS
6. November 2020 at 15:05
You wrote :
“Networks call states for no apparent reason. Fox News and AP called Arizona days ago, despite no clear evidence that Biden had won. It was just a guess. In contrast, it’s obvious that Biden has won Nevada and Pennsylvania, and yet the networks have not called those states. Networks should establish a consistent procedure and stick with it. I could do better than the clowns they hire.
If you don’t know, then don’t call it. How hard is that?
”
With that name calling (“clowns”) and believing you are smarter than people who do this for a living (“I could do better…”), you sound just like Trump.
6. November 2020 at 15:18
Scott,
that sounds good at first, there have been about 22 elections since then. But only 14 of them were elections with a challenger.
And in at least 6 of these elections, there were special effects. 1932 the Great Recession, 1940 and 1944 World War II, 2020 Covid-19.
In 1980 Reagan got 50,7% of the votes, that’s basically 51%, this is easily Biden’s dimensions. Reagan would have gotten up to 6% more without Anderson, so that’s the actual FDR memorial challenger election.
In 1992 there were similar special effects with Clinton, because Ross Perot won 19% of the votes.
@Randomize
The article does not explain much from my point of view. In Germany 30% of the votes are cast by absentee ballot. That is millions of votes. Soon it will be 40-50%.
It is strictly illegal to count the absentee ballots before the voting centers close. The counting is done by hand, counting machines are forbidden, but nevertheless the results are available on election day evening after a few hours.
Usually the counting takes about one hour, sometimes two or three hours at the most. I have often been an election worker, this is a question of logistics and manpower only, nothing more.
The article also doesn’t explain why most other states finish way earlier than PA.
I do not believe and hope that there is relevant election fraud. But the current US electoral system does not exactly inspire confidence either. And if you ever wanted to commit serious electoral fraud in a democratic country, you might count as slowly and as absurdly as some US states do.
6. November 2020 at 15:19
Michael, Even the mail-ins from red areas tend to be blue.
Jeremy, OK, I apologize for the “clown” comment; that was out of line. But not for the harsh criticism. Calling a presidential election is important, and states should not be called when they are very much in doubt. That’s just basic common sense.
6. November 2020 at 15:27
Appreciate the apology. In the case of Fox, after listening to their decision desk leader, it’s clear that their mistake was not that they called AZ when they still had doubt. Their mistake was that they didn’t have doubt — they misunderstood the remaining vote.
6. November 2020 at 20:24
Jeremy, OK.
I guess I’m frustrated with our entire election process, which is the laughing stock of the world. Next week there will still be millions of votes to count, for no good reason. The pace of vote counting (in some states, not all) is absurd. You get a batch of 200,000 votes, then 100,000, then 50,000, then 25,000, then 15,000, then 7000.
Other countries don’t have this problem.
7. November 2020 at 09:45
Totally agree with the slow counting. It’s so strange and in this day and age unacceptable. How was FL able to count so much more quickly? Did they actually learn something from 2000? Can other states learn from FL? Would national standards help? I am usually a federalist on most issues, but maybe there should be some minimum standards for counting speed and accuracy and some best practices for conducting elections/counting votes (e.g. type of ballot, machines used, manual recount triggers, etc). But can this dysfunctional US federal govt do it? I have very low confidence.