The Electoral College Enhances The Risk Of A Contested Election

One of the best arguments for the Electoral College, I would have thought, was that it reduced the risk and impact of a contested election. 

After all, you basically can't have and will never need a national recount in the Electoral College system, only recounts in a handful of states with close outcomes that also happen to be marginal states in terms of the required 270 electoral vote majority, a national recount seems awful, and two party systems naturally tend to develop coalitions in which each side has close to a 50% chance of winning.

But recent research casts some doubt on that intuition. As the Marginal Revolution Blog explains, citing recent social science analysis of the issue:

Starting from probabilistic simulations of likely presidential election outcomes that are similar to the output from election forecasting models, we calculate the likelihood of disputable, narrow outcomes under the Electoral College. The probability that the Electoral College is decided by 20,000 ballots or fewer in a single, pivotal state is greater than 1-in-10. Although it is possible in principle for either system to generate more risk of a disputable election outcome, in practice the Electoral College today is about 40 times as likely as a National Popular Vote to generate scenarios in which a small number of ballots in a pivotal voting unit determines the Presidency.

And note this, which explains a good deal of the debate and rationalizations — on both sides:

This disputed-election risk is asymmetric across political parties. It is about twice as likely that a Democrat’s (rather than Republican’s) Electoral College victory in a close election could be overturned by a judicial decision affecting less than 1,000, 5,000, or 10,000 ballots in a single, pivotal state.

That is from a new NBER working paper by Michael Geruso and Dean Spears. 

Further Analysis

Evaluating Contested Election Risk Well Is Possible

Election administration is essentially identical in any statewide election. The election administration issues would be identical, in any given state, under the Electoral College and under the popular vote system.

And, we have all sorts of experience with the risk of a contested election in an Electoral College system and in other statewide elections.

The analysis of a risk of a close election in a popular vote system should be much simpler. You add up the poplar vote totals from each state and you compare them. If they are sufficiently close, it is viable to mount a serious contest to the result. If the gap between the major candidate vote totals is significantly more than historical change in the outcome from each respective state's contested election outcomes, a contest isn't viable. If it the margin of victory is within the historically precedented margin of error of the vote count in 50 collective statewide elections and the DC election, then a contest is viable.

The overall national popular vote outcome is well predicted by a simple national polling average of likely voters, something we have exhaustive data upon. 

It is more cumbersome to get, but nothing but hard work separates a research from excellent quality historical data on the margin of error revealed in contested elections. 

And, the margin of error isn't very strongly sensitive to modest changes in the proportion of voters favoring each candidate in a particular state. Your margin of error estimate in a state gains an uncertainty of only about 10% with a five percentage point error in the proportion of voters who favor a particular candidate, which is a pretty big and unusual error (bigger than the polling errors in the swing states in 2016).

Four Potential Flaws In The Paper

I don't have access to the body text of the paper and my main concern about it are fourfold. 

First, it seems to be evaluating contest viability based upon raw vote differences in absolute terms when the appropriate measure is the percentage difference. A 20,000 vote gap nationwide out of 150,000,000 votes cast is worth contesting. A 20,000 vote gap out of 450,000 votes cast in New Hampshire to determine who won its electoral votes is not.

Second, it isn't clear if the study considers the impact of a contest in the respective cases properly. Even if it is rare, a nationwide recount in 50 different states and DC, each using different election administration systems, is a huge deal. A recount in two or three U.S. states that are potentially decisive in the Electoral College result is less disruptive and has far less moving parts to try to get right. Moreover, in the Electoral College, even if the state SOS certifies the wrong slate of delegates, once it is done, it is done. The Congress counts the ballots sent to it by the certified delegates than that is that, because the delegate votes (which have a discretionary element) and not the vote totals, is what Congress is counting. 

Third, it isn't obvious from the abstract that they are considering the fact that recount outcomes in Republican administered jurisdictions, and in Democratic administered jurisdictions, respectively, are likely to be strongly correlated, because the raise the same kinds of issues and administer their respective elections in similar ways.

Fourth, another important point is that in any given election, a contests will tend to happen with one method, or the other, but not both. This depends upon the national vote share of the candidates relative to each other, which, in turn, is largely a function of turnout.

In a year where a small number of votes in a few marginal states decide the Electoral College outcome, like 2000 or 2016, the popular vote is not going to be remotely close (the Democrat will lead by several million votes) and turnout will be moderately high.

In extremely high turnout years, Democrats win the Electoral College too without close races in the marginal state and contested elections are unlikely period. 

In a year where the popular vote is down to the wire, the Electoral College will be a clear win for the Republican with the marginal state having a several percentage point lead for the Republican, and turnout will be moderately low.

In extremely low turnout years, Republicans win the popular vote too and contested elections are likewise unlikely.

All other things being equal, a popular vote system should increase voters turnout relative to an Electoral College system, because it increases the relevance of the voting for both parties in states that are safe one way or the other for one party. So, a popular vote system itself, by increasing the likelihood of higher turnout than in an Electoral College system, reduces the likelihood of a contested popular vote outcome - at least until the candidates recalibrate their coalitions and messages to reflect the absence of the pro-Republican bias of the Electoral College.

Also, within that higher overall turnout range, turnout should be higher in years where national, general election polling is tight, and should be lower in years where national general election polling shows a clear favorite.

Does The Overall Conclusion Still Make Sense?

The hypothesis is still pretty plausible despite these concerns. Why?

It is almost guaranteed in an Electoral College system that one or several states will have an outcome that is close enough to contest, and candidates have a natural tendency to adjust their coalitions and approaches to match that tipping point. If any of those states are marginal or potentially marginal it makes sense to contest the election. 

But, in a popular vote scenario, the actual popular vote margin of victory has to be less than about 0.5 percentage points nationwide to be worth contesting, which isn't all that common historically (although the candidates would drift in their appeals towards that equilibrium if a popular vote system were adopted). 

Also, in a national popular vote system, there is a greater incentive for politicians nationally to insist on running a tighter ship in terms of election administration than they would now in a safe state where close partisan elections are rare in jurisdictions like DC or MA or AL or UT. This should reduce the margin of error that a recount or contest can produce relative to the status quo.

Not All Contested Elections Are Equally Important

One final thought.

The main axiom that justifies the electoral process in Presidential elections (although not the only one, even one party states find it useful to hold national Presidential elections) is that a Presidential election in which essentially all voters in the country participate is the best way to select the candidate who is most likely to act in the best interests of a majority of voters as those voters weight their concerns and priorities, among the candidates that are available for voters to choose from in the electoral process.

But there is no guarantee that the voters will make the right choice in any given electoral process. Some electoral processes may lead to the right choice more often than others, but the voters can make the wrong choice - not primarily because they are wrong about what their best interests are (although voters can get carried away with hype and misconceptions like anyone can), but because they are wrong about what the candidates will actually do if elected or what the consequences of those actions will be.

The closer the popular vote is in an election, the less important, in some abstract philosophical sense, it ought to be to have the election administration selection the correct winner. 

If 75,001,000 people vote for Candidate A and 75,000,000 people vote for Candidate B, the general public really has no idea who is the better candidate, perhaps because it is not possible to know with any accuracy. In an outcome like that, we are getting into questions of spurious precision in the electoral process, given the limited capacity of the voters to accurately determine which candidate will be better. 

The likelihood that the median voter made a bad decision in this scenario on the merits, for example, due to cognitive biases or campaign flourishes that don't reflect on likely actual performance on the job, far exceeds the risk that the vote count is masking an accurate determination by the voters of the best candidate based on something more than random chance.

If the popular vote is that close, either candidate is probably fine.

In contrast, the Electoral College, as it currently exists, has a built in three percentage point Republican bias, and the potential to be significantly worse than that in not particularly improbable possible outcomes. 

So, it is very hard for even a much higher precision determinations of the Electoral College winner under the Electoral College rules to outweigh that fact that the more precise the outcome is the more certain it is that it is systemically biased from choosing the right candidate.

Overall, the Electoral College is much more likely to choose the wrong candidate to be President. 

Supporting that conclusion, it is hard to conclude that the Electoral College made a better choice than the popular vote in any of the recent elections in which these measures reached difference conclusions. QED.