In casual conversation about politics, I see a growing assumption that all Coastal Liberals will reject the Electoral College: after all, our candidates win the popular vote. The “Hamilton Elector” attempts in 2016 frightened us. We remember Civics teachers telling us that the EC is there to prevent a demagogue from taking over the country. But we’re actually more scared about a demagogue using the EC, by influencing the smaller states to put a reactionary minority in charge.

I value the Electoral College as an anti-fragile element of our Republic’s design. This note is to explain why, and then to use that light to illuminate new angles on the standard arguments for and against the Electoral College. What are the important features of anti-fragility?

• Simple rules
• Decentralized evaluation
• Failure-damping layering
• Convey local failure signals up—drive adaptation
• Experience local failure locally
• Redundancy and overcompensation
• Trust local operators

Why have elections?

It is widely understood that the purpose of an election is to pick a winner—to pick someone to occupy an office. This is not so. Were it so, we have many easier ways to do so. The key contribution of an election is that it persuades everyone who is not the winner to sit down: that they have lost and are not in power. After an election, everyone—most especially the losing candidates and their supporters—agree on the victor. This is called legitimacy. Legitimacy prevents uprisings and civil wars.

Therefore, the most important threat to worry about to an election—the loss we clearly cannot accept—is a lack of legitimacy. We lose legitimacy when people believe that fraud has happened.

What else can go wrong? Someone could actually steal the election—there could be fraud that affects an outcome. Let’s focus most on the Presidency of the United States, the most powerful elected office in the world. Lots of people in the United States and outside it want to affect who has that office. But those inside the US don’t want to risk that legitimacy—it’s no good having the office if everyone knows you stole it! They’re only going to commit certain sorts of frauds. For some of those outside the country, breaking the legitimacy of our elections is a win all on its own.

How do our elections work?

Most of our states conduct an internal uniform popular vote. It’s organized by districts, but each district reports up a total number of votes for each candidate, and a total number of votes cast (as an error check). These totals are summed. All the electors go to whoever gets the most votes. Ties are handled comically: by a hand of poker, a coin flip, or letting some distinguished person choose. If the election fails to return a result, it is rescheduled and held again at moderate expense.

Some of our states vary a bit: they give some electors to the winner of the state popular vote, but distribute others to other candidates in proportion to the number of votes they received. Maine is notable here—but as it has only 3 electors, and leans strongly Democratic, one vote is up for grabs: the Democrats have a lock on two, and the Republicans might win one with an intensive campaign. This reduces national incentive to direct campaign resources towards Maine.

A few states have discussed assigning two electors to the winner of the state popular vote, and one elector to the winner of a vote in the heavily gerrymandered congressional districts. This is a terrible idea for all the reasons gerrymandering is terrible.

Evaluating the Electoral College

Against the criteria above:

The rules are simple: we can conduct an election in 2020 with pencil and paper carried on horseback. We’d rather not—but it would basically work. Unlike a Roberts’ Rules system, we don’t need a parliamentarian to interpret the rules and help us understand their interactions.

Elections are operated principally by volunteer retirees coordinated at the level of school districts. They handle questions like “can this person vote” with only delayed and sparse review. States manage their own machine and process selection in a nicely decentralized way.

The layers damp failures: a fraud in Oklahoma doesn’t affect the total, because we know Oklahoma’s electors will be Republican. Moreover, someone who does improperly win an election has little opportunity to turn that one win into others—you can’t gerrymander state borders.

But the failure is both signalled upwards by news media reporting, and experienced locally in terms of political fallout. Here in Massachusetts, William Galvin has been repeatedly re-elected Secretary of State running principally on a record of competence and integrity. States whose elections are poorly run replace those responsible quickly—even when run by governors with otherwise awful policies.

We use the on-the-ground system for not only Presidential elections, but for dog catchers and governors and the critical school board. It can support a primary in the summer of a year, a Presidential election that November, and a recall election that February. That part has moderate redundancy; we could do better, and if we ever have an illegitimate election we’ll have to do better. I suspect that will involve pencils and paper, then phone calls and hand delivery. Then and now, we’ll trust those local retirees to handle local problems: if a Russian agent shows up to mess with the election, they’re going to call local police, deal with the problem, and the rest of us will find out only much later.

What if there is election fraud?

In most years, very few states are close (see Wikipedia’s list of records). Minnesota will always vote Democratic. Utah will always vote Republican. In years where that’s not true, it’s also not close the other way—in 1964 and in 1984, a candidate won nearly every state.

A fraudster who wishes to steal an election has no reason to try to commit fraud in Minnesota or Utah: they’d need to create very many fraudulent ballots—and so are more likely to get caught—and assign them to many districts—and still nobody would look at exit polls and believe this outcome. Because of this, most states have no need for extraordinary election security. They have to worry about local actors committing local frauds in the race for governor or for school board—but we know who their electors are going to be.

If we remove the Electoral College, every state will have to defend against fraud in the same way that Ohio and Pennsylvania will now. Both parties will have to ramp up defenses and observations—and based on those observations, we should expect to see Bush v Gore legislation in tens of states, every time. What plagued Florida will visit any state where some local municipality used poor practices, because that municipality is contributing a key ten thousand votes to the totals.

But what about the popular vote?

Maybe it’s undemocratic to have representative electors rather than directly elect the President. But the electors do a pretty good job of modeling the outcome of a direct election. Illiterate news organizations run stories about who won the popular vote every four years—but we have not conducted a popular vote. News organizations saying we have badly misunderstand the facts on which they’re reporting. The best part is that these stories run right next to stories about low voter turnout, especially in states with predictable results.

In Massachusetts, most people are registered Democrats. Some are registered Republicans, and some are not enrolled. An inconsequential fraction are registered Green or Libertarian. The Green and Libertarian parties do not have meaningful primaries—at most one candidate runs—and do not affect election outcomes. Both the Republicans and the Democrats have incentive to vote in their party Primary. The Republicans can nudge the consensus of the party, and their delegates have been very effective at floor negotiations. The Democrats have a real say in who’s going to be on the ballot.

But when it’s time for the real election in November, both Republicans and Democrats know what the outcome is ahead of time. They can choose to stay home—and if the Governorship and both Senators are predictable, they overwhelmingly do stay home. That’s a few hundred thousand Republicans and millions of Democrats. The reverse is true in Oklahoma or South Dakota.

Across the country, tens of millions of people stay home who would plausibly come out to vote if their vote really mattered. In a real popular election, we should expect to see turnouts double our ordinary turnout of the last half century. When we try that experiment, then we can talk about who won the popular vote.

Until then, part of the duty of every citizen is to understand the arithmetic here and not mistakenly damage the legitimacy of our election process.

A modest proposal

On all the bases above, we should keep the Electoral College for the Presidency. We should strongly consider using it for state-wide offices, starting with the Governor and the Senators, using existing Congressional districts.

We might add a layer of hierarchy: states vote by internal ECs, one per congressional district. In Wyoming (1 rep), nothing changes. But now we have significant extra insulation against overvoting fraud.

As part of this bargain, of course, we would have to fix district boundary calculation to avoid gerrymandering. My hope is that Republicans want better election of state-level offices, and all statesmen want to fix gerrymandering—but the Republicans think it’s more expensive than the Democrats do, so need a feature they can take home.