The electoral college bias is defined as the difference between the Democratic margin of victory[1] in the tipping-point state minus the Democratic margin of victory in the national popular vote, in the 2024 presidential election. The tipping point state is defined as, if you were to order all states from highest to lowest by Democratic margin of victory, the one that would get the Democratic candidate[2] to 270 electoral votes[3].
For example, in the 2012 presidential election, the tipping point state was Colorado, which Obama won by 5.4%, while he won the national popular vote by 3.9%. Thus, the electoral college bias was D+1.5, and so this market would have resolved to "D+1 to D+2" for that election.
In the 2020 presidential election, the tipping point state was Wisconsin (because Pennsylvania alone would have gotten Biden to 269 electoral votes; Wisconsin was needed to get him to 270). He won Wisconsin by 0.6% and the national popular vote by 4.4%. Thus, the electoral college bias was R+3.8, and so this market would have resolved to "R+3 to R+4".
(Faithless electors are irrelevant to the resolution of this market.)
[1] Defined as the percentage of the electorate in the state that voted for the Democratic candidate, minus the percentage that voted for the Republican candidate.
[2] Presumably Harris, but if someone else somehow becomes the Democratic nominee then the question is about them.
[3] 270 rather than 269 because I expect that an exact electoral college tie would be resolved in Trump's favor.
1,000
3.00Nate Silver discussed five different scenarios for polling error in a recent post: https://www.natesilver.net/p/the-128-paths-to-the-white-house
Based on the five scenarios, we have five different ways the EC bias could unfold:
Polls are accurate. Based on the current Silver Bulletin polling average, this would mean about an R+2 bias (right now Harris's national lead is exactly 2 pts ahead of her lead in PA).
Polls underestimate Trump in the Rust Belt but are accurate in the Sun Belt. Worst-case scenario for Electoral College bias, probably around R+4 or even worse.
Polls underestimate Harris in the Sun Belt but are accurate in the Rust Belt. This could shrink the Electoral College bias if one of the sun Belt states ends up being the tipping point instead of a rust belt state. So probably something like R+1 to R+2.
Polls underestimate the Rust Belt/Sun Belt gap (racial depolarization scenario). In this scenario, the Rust Belt is bluer than we thought, so the EC bias would be significantly reduced. Probably Even to R+1.
Polls underestimate Trump in the Rust Belt and Harris in the Sun Belt. This would make the EC bias worse, maybe R+3 to R+4, similar to the 2020 election.