Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Avoiding conflict despite the Electoral College

November 3rd, 2020, 7:41am by Sam Wang

Don’t forget, the states in gold started counting their mail-in votes today or last night!

Above is the mode (single most likely combination) of this morning’s poll margins and uncertainties, converted to individual state probabilities. The margin doesn’t look that close, assuming the count proceeds in an orderly fashion in the coming days. How did we get here from 2016?

Last week at our Election Innovation Lab’s Friday Forum, an eminent colleague from the Politics department reminded me of a defense of the Electoral College that has been forgotten lately: it amplifies the popular vote margin to create the appearance of a mandate.

For two obvious reasons, the popular-vote/electoral-vote inversions of 2000 and 2016, this has been forgotten. But in this year’s unusual election, there is potential for this quirk to save us all from weeks of chaos. Here is why.

The core reason for this amplification is that winner-take-all races amplify the effects of popular-vote margins. Think about legislative and Congressional seats. If a candidate wins 51% of the vote, she wins 100% of her seat. If you add this up across many races, each with their own partisan lean, you end up with a translation of seats to votes that is nonlinear, looking like an S-shaped curve.

In this system, the popular-vote winner usually gets more seats. Typically a vote margin gets amplified by a factor of 2 or 3, depending on how much variation there is between districts, and the pattern of variation.

In the case of the Electoral College, the amplification factor this year is large, nearly 4. Biden’s ahead of Trump by around 9% in national opinion, and the map at the top of this post corresponds to a margin of (351-187)/538 = 35% of the electoral votes. If this holds up tonight as the early-reporting swing states of North Carolina, Georgia, and Florida report in, then the appearance of an electoral landslide might help to stop the hurricane of lawsuits and unrest that the Trump campaign seems to be ready to unleash.

You can see why this argument hasn’t been very hip lately. The popular-vote/electoral vote mismatches of 2000 and 2016 have pretty much killed this particular flavor of happy talk. I have calculated from historical election returns that over a wide range of political conditions, Presidential races with popular-vote margins closer than 3% will elect the second-place vote-getter in 1 out of 3 cases.

In short, the tradeoff is between the pleasing nature of a big blue or red map (recall Reagan 1984, or Clinton 1996, or Nixon 1972) and the risk of a game of Russian roulette whenever there is a close election.

This year, the Electoral College gives Trump a 2.5% advantage: a 2.5% popular loss would, on average, still give him even odds at winning the Electoral College. Biden is about 5-6% above that threshold. That’s what defines the Meta-Margin in the banner above. It matches very well a Biden national margin of 8% or so in the popular vote.

If the polling error in key states in 2016, 3 percentage points, were repeated across the board, the electoral map would still favor Biden:

However, this isn’t the biggest question this year. The question is what will happen as counts trickle in. Specifically, the states in gold below are only now starting to count their mail-in ballots. So to get a resolution of the Presidential race, we really need North Carolina, Georgia, and Florida to weigh in tonight.


The counterargument to this defense of the Electoral College is to consider the possibility of a national popular vote. If the election were held in that manner, Biden’s likely popular margin over Trump would be in the range of 13-14 million votes. That would also end the argument, and with much smaller risk of disputes. We also wouldn’t have this weird fascination with swing states, at least for the Presidency. Instead we could focus on Senate and local races, which are critical to the functioning of our democracy. So, like I said, two cheers for the Electoral College!

Tags: 2020 Election · President

22 Comments so far ↓

  • Scott

    Why the sudden plunge in the metamargin overnight?

    • Sam Wang

      I wouldn’t call it a “plunge.” Yesterday it was 5.4-5.6%, today it’s 5.3%. However, he answer to your question is that this tick might be caused by changes in states near the tipping point of reaching 270 electoral votes. I think the Pennsylvania median narrowed by about half a percentage point.

    • Matthew J. McIrvin

      Seems like a lot of perennially Republican-leading polls flooded the zone with numbers in the final hours, possibly to game the polling averages. Prof. Wang isn’t the type to just exclude them out of suspicion.

    • Matthew J. McIrvin

      (though I have to say, the shift puts the map closer to what my more pessimistic gut has been saying)

  • Tom Rutkowski

    Why does the map color PA darker blue than MI and WI if the median margin is smaller?

  • 538_Refugee

    I’m a little surprised a couple of states didn’t remain tossups on this map going into today.

    • Sam Wang

      There was one tossup: Georgia. I assigned it. That was hotheaded, sorry. Reverting to direct snapshot from 270toWin graphic of PEC probabilities.

  • paras

    The machine learning algos seem to be predicting a landslide
    55 45 pop vote
    372 166

    This one swung hard against Hilary in 2016, in the last three days. Seems to be swinging hard for Biden this time. Have you looked at this one Dr. Wang?

    • Nielen Stander

      Machine Learning cannot compensate for polling error or things that go wrong during voting or counting. As Sam said, that remains the unknown.

  • Abe Fisher

    Looking at how the EC outcome has or hasn’t reflected (or amplified) the popular outcome is maybe a little trickier than it seems. Up until roughly 100 years ago, the House of Representatives was increased after almost every decennial census. That would tend to push the EC towards alignment with (and amplification of) the popular vote. Since then, the two have diverged more and more, increasing the bonus to low population states.

    • Sam Wang

      That’s actually not true. Distortions in the conversation of votes to seats will always occur. It is a natural consequence of the standard deviation of vote share among states. A standard deviation of 15 points leads to an amplification factor of 2, no matter how many or how few seats there are. Basically the seats-votes curve is inescapable. See my Stanford Law Review article.

  • Nff

    Hi Sam:

    Reports on the ground from PA in regards to the 4 key counties – Westmoreland, Erie, Philadelphia & Chester – suggest PA could revert to a toss up as married white suburban women may be returning to 45.

    If PA goes red, is it over for Biden?

    Ty for your insight.

  • Electoral College Skeptic

    I don’t have strong feelings on the electoral college, though I generally favor the popular vote over the electoral college, at least in theory. In practice, a close race would be problematic, with recounts having to happen nationwide. Of course, we’ve landed people on the moon. Surely we can handle a nationwide voting system.

    Anyway, but I’m curious why amplifying the size of the win and mandate are positive aspects. Can’t people easily tell that even if someone won 60% of the electoral college but only 51% of the popular vote or let’s say just barely won the swing states to get to 60% of the electoral college, it was still a close race? Trump won in 2016 by 78,000 votes across 3 states, but the electoral college margin was much larger. But it was still a very close vote and if that were reported more, everyone would know it, so who cares that the electoral college margin was larger. I’m happy to wrong about this and am curious to hear your take.

    • Sam Wang

      I generally agree with you. See my article with Jacob Canter.

      Basically, the poli-sci idea is that the EC builds the feeling of a landslide. They built that argument from pre-2000 data. Recall the Reagan, Clinton, and Nixon wins as examples.

      However, as we now know, that comes with a very specific risk: a 1 in 3 chance of a failure in case of a close election. Is that feeling of a mandate worth the risk of having the second-place winner become president?

  • ArcticStones

    Not sure where to post this, but here are a couple of things I am pondering on Election Day:

    – How many mail-in ballots are still floating around in the slowed USPS systems?

    (NB. Approx. 95.6 million mail-in ballots were requested/sent; 66.2 million have been returned; 29.4 million are still outstanding – that is 31 percent yet to be returned.)

    – Are we seeing evidence of a successful Election Day Democratic GOTV effort in swing states? I’m especially curious about Florida, Georgia, Texas and Pennsylvania?

    – Significantly more woman than men have voted so far. Will this continue on Election Day? If so, given that women favor Biden by a 23-point margin, how much may this impact the election?

    • Sam Wang

      There is a news report of a federal judge’s order to USPS to speed ballots by whatever means necessary. This includes sweeping facilities for ballots and assigning them the status of Priority Mail.

      Polling stations are said to be pretty quiet. So many people have voted already!

    • ArcticStones

      “The NAACP, one of the groups suing the USPS, says the Postal Service did not comply with Judge Emmet Sullivan’s order. They have requested an immediate conference.”
      – John Kruzel, The Hill

  • Kevin Moore

    Regardless of the debates over categorizing and evaluating EC-generated distortion of the popular vote, there’s one absolute truth: The EC always distorts the will of the majority in one direction or the other. It doesn’t distort it as badly as the Senate, and because of the Senate, of SCOTUS, and not as badly as gerrymandering. But the inherent unfairness of all three marks the US experiment as disgracefully wrong and unfair if you believe in one human-vote vote.

  • Lorem

    I think a point that could bear more emphasis is that EC vs popular vote is not just about the counting – it’s also about the candidate incentives. If swing state votes are what determine victory, then presidential policies (and campaign styles, etc.) will be crafted to appeal to the median swing state voter, which could lead to significantly different law-making than policies crafted to appeal to the overall median voter.


    Wouldn’t a solid solution to the problem be providing Statehood to Puerto Rico, DC and even the Pacific Islands? Not only would you give disenfranchised Americans an actual voice, but you would add a few Electors and align the EC more with the popular vote–still giving it the multiplier effect and also make the Senate more fairly representative of the true US population.

Leave a Comment