Since the 2000 USA Presidential Election resulted in the first split between the Electoral College outcome and the popular vote in over a century, it has almost become a tradition every four years to speculate on the possibility of another popular vote/Electoral College split. A frequent topic is the small-state bias of the Electoral College. This is the result of how the Electoral College votes (EVs) for each state are calculated from the state’s Congressional delegation, which is composed of both a proportionally-apportioned component (Representatives in the House of Reps) and a component which is equal for all states (Senators, of which each state has two).
Thus we get the zany comparisons between the relative representation of Wyomingites and Californians, or perhaps in more nuanced analyses, discussion of how the Electoral College favours large battleground states. While both are significant issues, the small-state bias often garners much more of the attention than it deserves due to how relatively easy it is to explain.
Instead, the largest issue with the Electoral College is the fact that each state’s election tends to be winner-take-all – the electoral vote(s) of any jurisdiction all go to a single candidate (including the Congressional-District splits adopted by Maine and Nebraska). For example, none of the 4.5 million votes cast by Californians for Donald Trump in 2016 gave him a single EV, while the 4.6 million Trump voters in the 14 smallest electorates who voted for him awarded him 42 EVs, while the same holds for Clinton voters in strongly Republican states.
Those 4.5 million voters in California are effectively disenfranchised by the system, as their votes carry no weight in the assignment of EVs to any candidate. This also further entrenches a two-party system in American politics, for unless a third party candidate can win a plurality (the largest number) of votes in an electorate (any of 48 states + Maine/Nebraska CDs + DC), voting for them under the American system effectively tosses your vote.
A fix for this would be to switch the winner-take-all Electoral College elections for a proportional representation (PR) system, where a candidate’s share of the vote is roughly matched by their share of seats in each state + DC. Any proportional system would fix both of the above issues by giving representation to all voters in safe states as well as third-party voters, producing more representative results even without addressing small-state bias.
To further examine this, we look at USA elections since 2000 to examine the impacts of a proportional apportionment (eliminating the small-state bias) versus the adoption of proportional representation in all states + DC. For a quick refresher, here’s a list of all such elections:
(if you’re on a mobile device, scroll right for full data or turn your device landscape)
| Election | Democratic vote | Republican vote | Others vote | Democrat EVs | Republican EVs | Others EVs |
|---|---|---|---|---|---|---|
| 2000 | 48.4 | 47.9 | 3.7 | 267 | 271 | 0 |
| 2004 | 48.3 | 50.7 | 1 | 252 | 286 | 0 |
| 2008 | 52.9 | 45.7 | 1.4 | 365 | 173 | 0 |
| 2012 | 51.1 | 47.2 | 1.7 | 332 | 206 | 0 |
| 2016 | 48.2 | 46.1 | 5.7 | 232 | 306 | 0 |
| 2020 | 51.3 | 46.9 | 1.8 | 306 | 232 | 0 |
2020 election vote data is projected from the data available on 30/Nov/20.
Below are the methods we used to calculate EVs for each simulation:
For our proportional apportionment simulation, we first assigned each state + DC a single EV, then calculated a population quota [ (total population of 50 states + DC)/(remaining electors) ] which was used to calculate the number of EVs each state + DC “should” get. Where there were discrepancies in the EV total, largest remainder was used to add/remove EVs as necessary. Data for those calculations is available for download here.
For the proportional representation simulation, we used the EV counts for each state at each election. It was assumed that the proportional representation system used was single transferable vote, under which voters can rank candidates to allow their vote to flow from eliminated candidates to their next preference. A quota was calculated using the Droop formula, which is the most common formula used for STV calculations worldwide. EVs were then assigned based on the following criteria:
- All candidates who had whole quotas were assigned those EVs. Results were finalised in all states where candidate EV totals equalled their allocated EVs.
- Candidates with quota remainders greater than 0.8 were assigned an extra EV in that state, under the assumption that such candidates would be likely to gain the votes necessary to make up the difference from preferences. If 1 EV remained and there were 2 candidates with remainders >0.8, the EV was assigned to the candidate with the largest remainder.
- If there are very popular (>=10%) minor party/independent candidates who won above 0.5 quotas, they could be assigned an EV on the assumption that other minor/independent voters would likely preference them over the major parties (author’s judgement)
- If there were candidates who had a large gap between them and their next competitor (>0.4), they were assigned an EV.
- If there are very close results between major party candidates, minor party votes could be used to break the tie. Green voters were assumed to mostly preference Democrats while Libertarian voters were assumed to mostly preference Republicans.
- If there are still unassigned EVs and no easy way to discern the remaining candidates, largest remainder was used to assign such EVs.
As there have been nearly no ranked-choice voting elections in the US, these are the assumptions we used in an attempt to simulate an STV election. Do note that it is very likely that voter behaviour would have significantly shifted had proportional representation been in effect during these elections; we’re planning a piece on Duverger’s Law that examines various systems and such effects. Data for all of the proportional representation calculations is available here.
Removing small-state bias makes very little difference to overall results
(if you’re on a mobile device, scroll right for full data or turn your device landscape)
| Election | Popular vote | Actual EV | EVs, small-state bias removed | EVs, proportional representation |
|---|---|---|---|---|
| 2000 | 48.4 - 47.9 - 3.7 | 266 - 271 - 0 | 273 - 265 - 0 | 269 - 265 - 4 |
| 2004 | 48.3 - 50.7 - 1 | 252 - 286 - 0 | 256 - 282 - 0 | 258 - 280 - 0 |
| 2008 | 52.9 - 45.7 - 1.4 | 365 - 173 - 0 | 368 - 170 - 0 | 288 - 250 - 0 |
| 2012 | 51.1 - 47.2 - 1.7 | 332 - 206 - 0 | 338 - 200 - 0 | 274 - 264 - 0 |
| 2016 | 48.2 - 46.1 - 5.7 | 232 - 306 - 0 | 234 - 304 - 0 | 266 - 266 - 6 |
| 2020 | 51.3 - 46.9 - 1.8 | 306 - 232 - 0 | 309 - 229 - 0 | 274 - 264 - 0 |
2020 election vote data is projected from the data available on 30/Nov/20.
Some of the results presented in the above table should be taken with a grain of salt; in particular it’s very likely that the 2004 election would have played out very differently with a Democratic incumbent rather than a Republican thanks to the incumbency advantage enjoyed by USA presidents.
The key takeaway, however, is that removing small-state bias doesn’t seem to have much effect on the topline result, with the only election outcome it could have swung being the 2000 election (which is itself exceptional in terms of how close the Electoral College vote was). On average, small-state bias seems to have a net effect of switching 4 EVs to the Republican candidate; an effect which is completely swamped by the effects of winner-take-all.
Removing small-state-bias but leaving winner-takes-all still leaves the 1-5% of third-party voters completely disenfranchised, whereas under proportional representation Democrats and Republicans may be forced to negotiate with minor parties in close elections to get their preferred candidate elected (e.g. the 2016 election). Furthermore, proportional representation removes the Electoral College landslides-off-small-shares-of-the-vote common in winner-take-all by more accurately representing the voters in each electorate.
Although a lot of coverage on the Electoral College tends to focus on the times when it’s perceived to have “gotten it wrong”, i.e. resulted in the election of a candidate who did not win a plurality of votes, it’s also critical to consider the degree to which the Electoral College results in voters for minor parties not receiving any representation in the process of choosing a President. This is a concept known as misrepresentation, which broadly measures the difference between the share of votes won by each party/candidate and the share of the representation they are allocated by the electoral system. While there are many measures of this, here we use the Gallagher Index to quantify misrepresentation in the Electoral College (lower Gallagher Indices imply more representative results):
Abolishing winner-take-all would have a much larger impact than eliminating small-state bias
(if you’re on a mobile device, scroll right for full data or turn your device landscape)
| Election | Actual election outcome | Small state bias removed | Proportional representation |
|---|---|---|---|
| 2000 | 2.79 | 2.74 | 2.07 |
| 2004 | 1.99 | 1.29 | 0.96 |
| 2008 | 14.22 | 14.78 | 0.72 |
| 2012 | 9.82 | 10.93 | 1.33 |
| 2016 | 8.78 | 8.44 | 3.18 |
| 2020 | 4.85 | 5.43 | 1.73 |
| Avg. change | 0 | +0.2 | -5.3 |
Gallagher Index scores were only calculated for candidates who polled over 1% of the popular vote and/or would have won a seat under the proportional representation system used here. All calculations were performed as if there were no faithless electors.
2020 election vote data is projected from the data available on 30/Nov/20.
For context, when Canada considered reforming its electoral system, the recommendation of the committee formed to study the issue was that the government adopt a system which would keep the Gallagher Index to 5 or less. By this standard, even without eliminating the small-state bias of the Electoral College, adopting proportional representation in every state would massively improve the representativeness of the Electoral College.
Adopting state-based proportional representation vs a national popular vote
Of course, most advocates for reform of the Electoral College propose a national popular vote to directly elect the President. However, such a measure would face multiple issues with regards to its implementation.
Firstly, there is the question of how the will of the electorate is to be best assessed in a national vote. While the simplest, most intuitive system of awarding the Presidency to the candidate with the most votes does sound appealing, a large third party vote would complicate it. Should a candidate, such as a Democrat with 40% of the vote, be considered the legitimate winner if they won 40% while the Republican won 39% and a 3rd party won 21%? Such scenarios shouldn’t be considered too outlandish given independent Ross Perot won 18.9% of the vote in the 1992 election under the winner-take-all system – if more people feel like independents and minor party candidates can win, they are likely to attract more votes. What if the third party was a right-leaning party whose voters would have preferred the Republican over the Democrat in a one-on-one race? Or what if the third party was a left-leaning party whose voters, left to their natural inclinations would have preferred the Democrat but the candidate would prefer the Republican?
(While there are alternatives such as two-round voting (where the top-two candidates from the first election proceed to a runoff) or ranked-choice voting (which enables calculation of a runoff vote from a single round), each option has its own tradeoffs in terms of complexity versus collecting more information about the will of the voters.)
Secondly, and more crucially, the only way to abolish the Electoral College is through constitutional amendment. This is highly unlikely to happen as a supermajority (two-thirds) in both the House of Reps and the Senate is required for such an amendment to even be brought to the states for voting upon. Then, for the amendment to pass, three-fourths, or 38 states, need to vote for it, meaning that the objections of just the 13 smallest states can scuttle the vote. Some of these states could also join with swing states, who would see their influence drastically diminish with the adoption of this amendment, to block the amendment.
Furthermore, in the current environment where opinions on the Electoral College are highly polarised by party identification and the party opposed to abolition – the Republicans – controlling 29 state legislatures, it is extremely unlikely such an amendment would be adopted. While there are attempts to circumvent the Electoral College through the National Popular Vote Interstate Compact (NPVIC) instead of a constitutional amendment, there is the possibility that such a compact would be illegal without the consent of Congress,
Additionally, in contrast to a constitutional amendment, the NPVIC is inherently unstable as changes in the state legislature or challenges from citizens can easily overturn a state’s participation in the NPVIC, making it much harder for the NPVIC to reach 270 EVs and take effect. Finally, as the NPVIC proposes for member states to give their EVs to the presidential candidate with the largest popular vote tally, it runs into the aforementioned issues surrounding uncertainty regarding the will of the electorate in any election with a significant third-party presence.
In contrast, state legislatures generally have the power to adopt different methods of selecting electors, as the Congressional District splits in Maine and Nebraska demonstrate. Hence, proportional representation is a plausible reform which can be adopted relatively rapidly. Unlike the Congressional District method, which strongly favours Republicans, proportional representation does not favour either major party and thus can be plausibly adopted as a bipartisan reform.
In particular, one way of increasing adoption of proportional representation might be for small, safe Democrat and Republican states to agree to simultaneously adopt PR. For example, Vermont and Wyoming, both of whom have 3 EVs, could adopt a law to allocate their EVs by proportional representation as long as the other state did so. In such safe states, the Democrat/Republican respectively would basically always win 2 of 3 EVs barring a landslide in either direction (in which case EV-by-EV calculations are highly unlikely to matter anyways). For either candidate to win all 3 EVs, the losing major party candidate will have to poll less than half a quota, or under about 12.5% of the vote, something which has only happened once in any state nearly a century ago during a massive landslide (1936, FDR v Landon, South Carolina and Mississippi) in electorates where African-Americans suffered from severe vote suppression. If the PR system used was ranked-choice, they would then have to fail to nab any preferences from third-party voters in order to be excluded. Such an agreement could be gradually expanded to larger and larger states (e.g. Rhode Island/Idaho, New Mexico/West Virginia) as a means of ensuring that every vote in such states can contribute to the election of the President as well as a method of encouraging Presidential candidates and parties to pay greater attention to the concerns of voters in safe states.
Potential issues with proportional representation
Proportional representation does have its flaws, however. Due to the massive disparity in EVs between states like Texas and North Dakota, it is easier to win an additional EV by campaigning in larger states, where the share of the vote required per EV is lower than in smaller states. For example, in California, a candidate can win EVs with just 1.8% of the vote while in 3-EV states, a candidate requires about 25% of the vote to be elected. Note that this does not necessarily mean that the number of votes required to win an EV is lower; in fact due to the small-state bias of the Electoral College more votes are required per EV in larger states (e.g. the 2016 California quota is 253k votes) than in small states (e.g. the 2016 Wyoming quota is 64k votes).
Unlike the misfires of the winner-takes-all Electoral College, however, this is not an outright failure of misrepresentation, but rather depends on the philosophy one prefers when approaching the process of electing a government. If you prefer an election process which requires the inputs of as wide a spectrum of parties as possible, even those who won very tiny shares of the vote, allowing electors to be voted in even with tiny shares of the vote is not a bug but a feature that should be expanded.
On the other hand, if you prefer a more majoritarian (an often misused word in USA politics given that a majority is not required to win in most elections, Maine and Alaska excepted) approach to governance, then the idea that just 0.8% of the electorate can hold much larger parties hostage In our simulation of a PR 2000 presidential election, Gore would have won 269 EVs, just one short of a majority. Nader would have won 4 EVs and could have instructed his elector in Massachusetts, which does not penalise faithless electors, to vote for Gore to give him an EC majority. For Gore to be denied a majority, he had to fail to win all four of Nader’s EVs, and the total Nader vote in those states (CA, MA, NY) comes to about 0.8% of the total electorate. by means of how they direct their electors to vote will seem absurd and an issue which has to be remedied before proportional representation is adopted.
Nonetheless, there are solutions to this issue in larger states. By breaking the state up into smaller electorates which have a certain number of electors each e.g. California could split into 11 electorates of 5 electors each, which would increase the vote share quota from 1.8% to 16.7% per electorate. or by setting a minimum required vote share to win electors Commonly used in many proportional-representation systems, usually set at about 3 to 5%, the share of votes required to win an EV can be increased to prevent fringe parties from garnering undue power in the process of electing a President. Furthermore, if a ranked ballot form of proportional representation is adopted, votes for candidates who do not obtain sufficient votes for an EV on their own could go to the voters’ next preference, allowing all voters’ votes to count towards the result regardless of their first choice.
Hence, adopting proportional representation will produce a much more democratic result representative of the breadth of opinion in the USA electorate, solving the largest source of misrepresentation in USA Presidential elections.