Methods

How this works: We measure a state's relative margin as the difference between its presidential margin and the national presidential margin. A state with a relative margin of +5% is 5 points more Democratic than the nation as a whole, while a state with -3% is 3 points more Republican than the nation. By shifting the national presidential margin (PV) we can estimate how many electoral votes each party would win if the national popular vote were different. The EV bar above shows the estimated electoral vote split for the selected PV.

We only capped the maximum PV shift to ±50% to allow for particularly unrealistic scenarios (despite the days of a 20-point landslide being long gone). Since 1968, the largest PV margin was 23.1% in 1972, even Reagan never cracked 20% in his historic 1984 landslide. Since 1984, no candidate has surpassed a 10% margin. But if you want to see what a D+44.3 margin might look like, be my guest!

Note: This is a simplified model that assumes uniform swing across all states and does not account for factors like turnout changes, demographic shifts, or unique state-level dynamics. It is intended for illustrative purposes only. The assumption of a uniform swing is a significant simplification, but is slightly more reasonable for modern elections where we have a relatively common national zeitgeist.

Note on 1968: A few states had a strong showing by third-party candidate George Wallace, which complicates the uniform swing assumption. We assume the national swing applies purely to the D and R votes, and that Wallace's vote share remains constant. Thus, some of these states actually have two tipping points: usually pushing a Wallace win into a D/R win, but in the case of TN, pushing an R win into a Wallace win and then into a D win.
While a few states like Alabama and Mississippi are solidly Wallace territory (no national swing could change his plurality there), and other states like GA and LA require massive 30-50+ swings, other states like AR and TN have more reasonable tipping points.

Note on Alabama: Alabama is a complete mess. It's impossible to count it as a simple D/R contest in 1960 because of unpledged electors. We have opted to use the way Wikipedia counts them here. For testing different scenarios, we choose to allot electoral votes as 5D/6O unless Nixon wins, in which case all 11 go to R. In 1948, Truman was not even on the ballot, so we count Strom Thurmond as the de facto Democrat (but still color the state yellow unless it flips).

Flip scenarios are computed by solving a knapsack problem to find the minimum number of popular votes needed to flip enough states to change the electoral college outcome (either to produce a new winner or to simply break the majority the original winner had). We can choose the solution based on the minimum number of votes needed to flip or the minimum total percent of state margins flipped. For example: in 2024, if Harris had won AZ, GA, and NC, then it would have been an Electoral College tie. In each of those states, we'd only be flipping 1-3% of the votes. But that would have actually required more votes total (242,768) than flipping NV, MI, PA, and AK (only 145,408, which is nearly 100 thousand fewer, but the 22,217 votes in AK would have been 6.57% of all votes cast in that state), so we show both solutions under different metrics.