Big news doesn't come evenly, one news event per day. Instead, most days have no real news, but every once in awhile, there's a giant typhoon, with big news leading into more big news leading into more big news. The 2020 election sparked one such typhoon. So, this two-part blog post will be about the myriad ways that this election shaped my view of politics.
In Part 1, I'll talk about how close the 2020 election actually was, the new highs for Electoral College bias, and the varied accuracy of my predictions on the election. Then in Part 2, I'll move to the usefulness of electability, the value of intuition versus uncertainty (and why FiveThirtyEight is so great), as well a segment on why exactly Democrats didn't win in a landslide (featuring discussion of dark-matter Republicans). Let's get right to it!
The Elephant in the Room
There's one big obvious thing about the 2020 election that's pretty weird (well, a lot of big obvious things, but one of them): Donald Trump got scarily close to winning. Joe Biden had a a huge lead in the polls, Trump mishandled a pandemic leading to many deaths, he had enormous personal flaws and many debate gaps, and so many other things which I won't even try to list because I'm sure everyone reading this is already on the same page. And what was the margin?
Well, Biden ultimately won with 306 electoral votes. The closest state by votes was Arizona, which Biden won by just 0.31%, or 10,457 votes. If a majority of those people–5,229–flipped, that would have allowed Trump to win by one vote. Georgia was next-closest, requiring 5,890 flips. And finally, Wisconsin third, needing 10,342 flips. That's enough to create a 269-269 electoral college tie which (given the fact that states themselves are gerrymandered towards Republicans and stuff, and maybe here you could also talk about the Supreme Court's conservative bias) would have presumably resulted in a Trump win. But Nebraska's 2nd congressional district (Nebraska and Maine do electoral votes by district) would have changed from Biden to Trump with only 11,046 swaps, so let's flip that too. And now suddenly:
(I should probably be using the 435 Project here, but 270toWin's easier for now.)
Without context, the 32,507 votes I changed may not seem like much. But turnout in the 2020 election was record high: 158,383,403 people voted. Well, we can divide. The people whose votes ultimately decided the 2020 presidential election represent just over 0.02% of voters. One in every five thousand. That's an astonishingly low statistic. A rounding error. Purely from a statistical perspective, the 2020 election was essentially a tie.
The people whose votes ultimately decided the 2020 presidential election represent just over 0.02% of voters. One in every five thousand.
For comparison, that statistic in 2016 was 0.03%–making the 2020 election even closer than the 2016 election. (In 2012, by the way, it's 0.35%; in 2000, 0.0002%.) This comes despite Biden winning 51.3% of the popular vote compared to 2016-Trump's 46.1%. Which brings me to...
The Worst Electoral College Result Since 1948
I'm not sure if it's a tradition exactly, but for each of my birthdays since I started Chromatic Conflux, I've done some post related to the Electoral College: the two-parter back in 2019, and the follow-up post in 2020. I did make an Electoral College spreadsheet for those, but I no longer understand what the formulas do and am slightly intimidated by it, so I'll just cheat and copy FiveThirtyEight's chart:
As you can see, the gap between the result in the tipping-point state (basically the state that put the winning candidate over the top) was mostly negligible between 1952 and 2004, though it was pretty sizable in 1948 (when it advantaged Thomas Dewey over the ultimately-victorious Harry Truman). That, of course, includes 2000, where the slightest of Electoral College edges ultimately flipped the outcome, showing that even minor discrepancies can matter. Democrats had non-negligible advantages in 2008 and 2012, but they didn't end up mattering.
And then, of course, came 2016, when the particulars of the election caused the Democratic Party to run up the vote in blue states like California at the expense of swing states. 2020 only exacerbated this trend, further highlighting the Electoral College's bias. I don't think I need to go over this–hopefully everyone can see how blatantly undemocratic it is to give all the deciding power to people in swing states over people in safe states.
Will this trend get worse in 2024? I tend to think it won't, just because of regression to the mean, but I don't think it'll improve much at all. Somewhere in the R+1 to R+4 range probably. It's dangerous because you can't expect a political party, even an obviously correct political party, to consistently win by at least 1 to 4%. You have to count narrow wins as wins, and the Electoral College doesn't!
However, like so many things, I don't control the Electoral College. What I do control, though, are my predictions of the future. And...let's take a look at how those turned out.
I Was Wrong
I was so sure it would be a blowout. In "Should Presidents Be Elected?", a July post which wasn't even about the 2020 election, I went out of my way to describe Trump as "the president immediately before Joe Biden." In fact, prior to the election, I suspected that the polling average–which showed the Democratic nominee ahead by around 8.4%–was more likely to be low than high. My intuition told me that the outlier ABC News/Wall Street Journal poll of Wisconsin with Biden up 17 might not be too far off the mark.
In "Should Presidents Be Elected," a post which wasn't even about the 2020 election, I want out of my way to describe Trump as "the president immediately before Joe Biden."
The theory made sense–after polls underestimated Trump's support in the 2016 election (in the Midwest), they wouldn't want to be accused of the same thing again, so they'd overcorrect. After all, it's some sort of historical law that people only care whether polls are "right" in the sense that they have the winner ahead. No one cares about the margin. So why not show a close race just to be safe?
Of course, my intuition turned out to be total garbage. The polls overshot Biden's support, not undershot it, in a polling error greater than the much-publicized one in 2016. That Wisconsin poll? Sixteen (16) percentage points off the mark. I was very wrong, and I talked about my opinions to a lot of people. I apologize.
And yes, I'm cherry-picking...a bit. And I'm strawmanning Past Jacob...a bit. I definitely wasn't going around saying Trump had no chance of coming close to a victory. But I was kind of implying that, and my notoriously-bad-at-conceptualizing-probabilities human brain was totally on board.
I definitely wasn't going around saying Trump had no chance of coming close to a victory. But I was kind of implying that, and my notoriously-bad-at-conceptualizing-probabilities human brain was totally on board.
My Senate and Gubernatorial Predictions
In any case, while I didn't give the presidential race official percentages, probably the most impartial look comes in my Senate and Gubernatorial Race Guides. Let's start with the Senate ones–here are all the Senate races by their categories (according to me) with the ultimate winner bolded:
"Safely Democratic"
Delaware: Sen. Chris Coons (D)
Illinois: Sen. Dick Durbin (D)
Oregon: Sen. Jeff Merkley (D)
New Jersey: Sen. Cory Booker (D)
Rhode Island: Sen. Jack Reed (D)
Virginia: Sen. Mark Warner (D)
Massachusetts: Sen. Ed Markey/Joe Kennedy III (D)
"Safely Republican"
Alaska: Sen. Dan Sullivan (R)
Arkansas: Sen. Tom Cotton (R)
Idaho: Sen. Jim Risch (R)
Louisiana: Sen. Bill Cassidy (R)
Mississippi: Sen. Cindy Hyde-Smith (R)
Nebraska: Sen. Ben Sasse (R)
Oklahoma: Sen. Jim Inhofe (R)
South Carolina: Sen. Lindsey Graham (R)
South Dakota: Sen. Mike Rounds (R)
West Virginia: Sen. Shelley Moore Capito (R)
Tennessee: Bill Hagerty (R)
Wyoming: Cynthia Lummis (R)
Quick note here–while all of my "safe" forecasts did come through, I think Lindsey Graham's seat in South Carolina was not safe in retrospect. In fact, there was some evidence that Jaime Harrison would've been well-positioned to win in the Democratic landslide scenario. Ultimately, Graham was reelected without much trouble, but I think it was wrong to give him a 100% chance. (One of my two real errors.) Perhaps you could claim something similar about Dan Sullivan in Alaska, but I think his reelection being guaranteed was less of a stretch.
"Likely Democratic"
Michigan: Sen. Gary Peters (D) vs. John James (R)
Minnesota: Sen. Tina Smith (D) vs. Jason Lewis (R)
New Hampshire: Sen. Jeanne Shaheen (D) vs. Don Bolduc/Bill O'Brien (R)
New Mexico: Ben Ray Luján (D) vs. Elisa Martinez/Mark Ronchetti/Mick Rich (R)
All the "likely" Democratic wins did ultimately come through without much drama; the ratings were fine. However, Jeanne Shaheen's opponent in New Hampshire ended up being Corky Messner, who I didn't list–the other thing I'd consider now to be a real error.
"Likely Republican"
Georgia (normal): Sen. David Perdue (R) vs. Jon Ossoff/Teresa Tomlinson (D)
Georgia (special): Sen. Kelly Loeffler/Doug Collins (R) vs. Raphael Warnock (D)
Iowa: Sen. Joni Ernst (R) vs. Theresa Greenfield/Kimberly Graham/Eddie Mauro (D)
Kansas: Kris Kobach/Roger Marshall (R) vs. Barbara Collier (D)
Kentucky: Sen. Mitch McConnell (R) vs. Amy McGrath (D)
Texas: Sen. John Cornyn (R) vs. MJ Hegar/Royce West (D)
So the only real surprises here come in the form of the Georgia runoffs, for which I am so unbelievably glad I was wrong. The Ossoff vs. Perdue race is particularly interesting, since Perdue literally got 49.7% to Ossoff's 47.9%, narrowly avoiding an instant win by under half a percent (the Libertarian candidate received the remaining 2.3%.) So you'd think that as long as Ossoff couldn't suddenly get virtually all of the Libertarian vote–a bit of a stretch–he'd be doomed. However, Perdue ultimately lost ground in the runoff due to a decrease in Republican turnout, causing Democrats Ossoff and Warnock to win in perhaps one of the most important political upsets in modern history! (The only recent one that comes to mind as more important is, well, you know.)
But anyway, I'd say these were classified correctly given the information available at the time.
"Leans Democratic"
Colorado: John Hickenlooper (D) vs. Sen. Cory Gardner
"Leans Republican"
Alabama: Jeff Sessions/Tommy Tuberville (R) vs. Sen. Doug Jones (D)
Montana: Sen. Steve Daines (R) vs. Steve Bullock (D)
"Tossup"
Arizona: Sen. Martha McSally (R) vs. Mark Kelly (D)
Maine: Sen. Susan Collins (R) vs. Sara Gideon/Betsy Sweet (D)
North Carolina: Sen. Thom Tillis (R) vs. Cal Cunningham (D)
Tossups are tossups. These ultimately did decide control of the Senate, and there wasn't much more I could've said. It is worth noting that Martha McSally has now lost two Senate elections in Arizona in two years.
But anyway, I'd say these were classified correctly given the information available at the time.
Overall, I'm pretty proud of how I did! Now, let's do the governor's elections.
"Safely Democratic"
Delaware: Gov. John Carney (D)
Washington: Gov. Jay Inslee (D)
"Safely Republican"
Indiana: Gov. Eric Holcomb (R)
Missouri: Gov. Mike Parson (R)
North Dakota: Gov. Doug Burgum (R)
Utah: Jon Huntsman/Spencer Cox (R)
West Virginia: Gov. Jim Justice (R)
"Likely Republican"
New Hampshire: Gov. Chris Sununu (R) vs. Dan Feltes/Andre Volinsky (D)
Vermont: Gov. Phil Scott (R) vs. David Zuckerman/Rebecca Holcombe (D)
"Leans Democratic"
North Carolina: Gov. Roy Cooper (D) vs. Dan Forest (R)
"Leans Republican"
Montana: Greg Gianforte/Tim Fox (R) vs. Mike Cooney/Whitney Williams (D)
In contrast to the Senate races, there were no surprises in the 2020 gubernatorial elections and no tossups. So I "called" each race perfectly, but it was pretty easy to do so.
But it gets more complicated. Closer to October and November, I didn't blog about it, but it did start appearing more likely that Democrats would win the tossups. In Maine, Susan Collins was behind in pretty much every poll, and Cunningham led convincingly in North Carolina. So in some states, polls were super wrong. But for other states, like Georgia, the polling average was almost exactly on the mark–despite polling error elsewhere.
It's near-impossible to really assess predictions like these with only one data point: what actually occurred.
In other words, my opinions about the election got less correct as 2020 went on. Which is weird, and maybe partially attributable to the pandemic making it look like Democrats would do better than they actually did. Then again, maybe Democrats would've done even worse without the pandemic. It's near-impossible to really assess predictions like these with only one data point: what actually occurred. The butterfly effect suggests that even apparently small differences can lead to big changes. And what about big differences?
I've...written a lot here about the 2020 election, but I'm nowhere near done. I'll see you next week in Part 2 to the virtues of electability, the issue with intuition, and the invisible Republicans among us.
–beautifulthorns
Comments