Winner take all

Born in a time when Congress was seen as the more important branch of government, the Electoral College, as implemented today, is a fundamentally undemocratic institution. The most anti-democracy aspect of the Electoral College isn’t the apportionment of electors (the “small state bias”), or the loyalty pledges and laws, but the winner-take-all allocation used by nearly every state. Only Maine and Nebraska divide their electors proportionally (though by district so the allocation isn’t purely proportional). It creates an idea that California as a whole voted for the Democratic candidate or the entire state of Texas voted for the Republican. This suppresses turnout, inhibits viable third parties, and reinforces polarization. It also means the candidates have little incentive to campaign in states they believe safe, at a risk, as Hillary Clinton found out.


Only 38,872 votes needed to go the other way for Clinton to have won the 2016 election. That’s 0.0121% of the population, 0.0168% of eligible voters (231,556,622), 0.0280% of votes cast (138,884,643), by all measures a rounding error (if we inexplicably rounded votes). Only 0.0280% of the people who participated in the election decided the outcome. Not 2.8% mis-divided, but 0.0280%. And more importantly, these votes were only in three states: Pennsylvania, Michigan, and Wisconsin. Meanwhile, 40.45%, over 54.8 million votes, were cast for both candidates in a state that went the other way, more than 1,400 times the number of decisive votes.

Of course, it’s not true that only those 38,872 votes mattered and the remaining 138,845,771 votes cast in the election aren’t important. Every single voter has a say in this step of the process. “Safe” states are only safe because of turnout.

Looking at how people voted across the country, some facts stand out:

  1. More people voted for Trump in California (3,916,209) than any other state except Texas & Florida. That’s more than 14 states combined: Vermont, Hawaii, Wyoming, Rhode Island, Delaware, North Dakota, South Dakota, Montana, New Mexico, Maine, New Hampshire, Idaho, Utah, and Nebraska. For some interesting context, this is also more than the population of the United States at the time the Electoral College was created, approximately 3.5 million people.

  2. More people voted for Clinton in Florida (4,485,745) than any other state except California. That’s more than 17 states combined: Wyoming, North Dakota, South Dakota, Montana, Vermont, West Virginia, Idaho, Delaware, Rhode Island, Hawaii, Nebraska, Utah, New Hampshire, Maine, Arkansas, New Mexico, and Kansas.

  3. The number of votes that gave the election to Trump is less than the number of people in “solid blue” Manhattan who voted for Trump. And it’s only a few thousand more than the number of Trump voters in The Bronx.

But even Manhattan isn’t so purely Democrat-voting. A full 10% of its voters cast ballots for the Republican candidate. The vast majority of counties in the US had at least 10% of votes go to the losing candidate of their state. This map shows the 54,817,232 votes that are effectively “wasted”, in the sense that they have no representation in the electoral college:

The 2016 US Presidential election results. States are colored by the candidate that lost the state (red = Trump, blue = Clinton), and the counties are shaded by the percentage of their votes that went for the losing candidate. Maine and Nebraska are excluded as they allocate electors proportionally by district.

The urban–rural divide is very real and growing, but it’s not absolute. There are still plenty of Republican voters in urban counties, and plenty of Democratic voters in rural counties. No county had more than 86% of its votes go to the losing candidate in their state. The most lopsided counties tend to be smaller.

The 2016 US Presidential election results by county. Each county is colored by the candidate that lost the state (red = Trump, blue = Clinton). The x-axis is the percent of votes in the county for the losing candidate, while the y-axis is the count of total votes in the county, on a log scale.


  • America is substantially more “purple” than people talk or act
  • Winner-take-all is foolish and severely undemocratic
  • Not voting is surrender, but the winner-take-all method helps to encourage this apathy.*

What can be done?

The winner-take-all approach is clearly contrary to the idea of representative democracy. Alexander Hamilton and James Madison assumed electors would be chosen on a proportional, per-district basis and Hamilton drafted an amendment after states did otherwise.

Changing apportionment of electors, or abolishing the Electoral College entirely, requires a constitutional amendment. Not impossible, but a substantial hurdle. Elector allocation_, however, is entirely up to the states on their own. There are already efforts to make allocation proportional, such as the National Popular Vote Interstate Compact. Allocating based on _national popular vote would outright undermine the Electoral College for the sake of reducing the small-state bias, but is only properly democratic if enough states cooperate. Allocating based on congressional district, as was intended, is susceptible to gerrymandering in a way unimaginable in the 18th Century. Allocating based on state popular vote more directly reflects the will within each state, but preserves the small-state bias derived from a compromise to preserve political power for those who enslaved people. Either way, surely the 620,825 people in Los Angeles who voted Republican, or the 458,845 people who voted Democratic in Dallas, want their voices heard.

*However, it’s important to acknowledge reasons people from historically and currently oppressed groups may choose to not participate in a system that oppresses them. Likewise, many, typically poor or non-white, are prevented from voting because of voter restrictions that specifically target them. These are both clearly not situations of “surrender”.

Maps generated with D3.js, using the TopoJSON US Atlas. County-level voting data is from via Tony McGovern.