Adrian Kavanagh, 18th November 2016 – latest update: 8th February 2017
As the dust settles following the conclusion of the 2016 USA Presidential Election campaign, there are two striking patterns, or trends, which point towards the high degree of disproportionality associated with the Electoral College system, as with other “first past the post” electoral systems:
- Donald Trump can be seen to have won a very clear 306-232 victory in the Electoral College
- Hillary Clinton emerged as the candidate with the highest number of “popular votes”. She held a lead of nearly three million votes over Donald Trump in the popular vote (over 2.8 million votes based on the latest updates)
Rather than getting into a pointless debate about the “unfairness” of the system, as an electoral geographer my argument is that you have to take the rules of the electoral game/system and shape your campaign to put these to your advantage. This is something that the Trump/Republican campaign clearly did in 2016. Instead, this post will try to tease out why there was such a mismatch between the percentage of the electoral college votes won by the two candidates and the percentage of the popular vote won by them. It is worth noting that British electoral geographers, Ron Johnston, David Rossiter and Charles Pattie, have published some very useful/interesting work on the levels of disproportionality evident at the 2000 and 2004 US Presidential elections. This work (as well as their work on the British electoral system) is well worth a read if you would like more in-depth analysis/discussion of the issue of disproportionality in “first past the post” electoral systems, such as those in the USA and the United Kingdom.
First, to address the claim made by some commentators, who have referred to the 2016 election as being the “least proportional” in terms of the history of US presidential elections. A review of the facts and figures relating only to the most recent presidential elections show that this is patently not the case.
|Year||% Popular Vote||% Electoral College votes|
Table 1: Comparison of the percentage of the popular vote and the percentage of the electoral college vote won by the successful presidential election candidates at election between 1980 and 2016
As Table 1 shows, the winning candidate (since 1980, at least) has usually enjoyed some degree of a “seat bonus” – i.e. they have won a higher share of the electoral college vote than their percentage share of the popular vote. The figures in Table 1 shows that the 2016 election compares relatively quite favourably with most recent elections in terms of the relative mismatch between electoral college vote and popular vote levels. In 2016, Trump will get a “seat bonus” of just over eleven percent. This is similar in scope to the bonus enjoyed by Barack Obama in the 2012 contest, but is notably smaller than the bonus enjoyed by Obama at the 2008 election. Bill Clinton enjoyed much larger “seat bonuses” than that of Trump in 2016 when he won the 1992 and 1996 elections, and this was also the case for the elections won by Ronald Reagan (1980, 1984) and George Bush (1988). Ironically, the most proportional of the elections held over the past 36 years were those won by George W. Bush in 2000 and 2004. The “most proportional” election of all was the 2000 contest, where Bush only enjoyed a “seat bonus” of 2.5%, although this was enough to ensure that he won the electoral college, despite failing to win the popular vote.
So why do we get disproportionality in US Presidential elections. As noted already, the work of Johnston, Rossiter and Pattie (2005, 2006B) sheds much detailed light on this in terms of “decomposing” the different factors that resulted in bias at the 2000 and 2004 contests. This concept of bias is an interesting one. Johnston, Rossiter and Pattie (2005, 2006B) explain bias as being as the difference between two parties (or candidates) in the number of seats (or electoral college votes( that they would obtain if they were to win the same share of the votes cast/the national vote. This post will apply some of their ideas to the November 8th election, but details on these articles can be also found at the bottom of this post.
It is worth noting that the issues to do with gerrymandering, associated with House of Congress elections, do not apply to the presidential election contests as the electoral boundaries for these contests are largely “set in stone” – the boundaries of the fifty US states (and the District of Columbia) are always the electoral boundaries for the presidential election contests. The notable exception to this rule applies to the congressional district boundaries in the states of Maine and Nebraska, due to the Maine-Nebraska system in which two of these states’ electoral college votes are given to the candidate who win the most votes in that state, but the other two (in Maine)/three (in Nebraska) electoral college votes are assigned to the candidate who win the most votes in those states’ congressional districts. In 2016, for instance, Hillary Clinton won the state of Maine, but Donald Trump took one of Maine’s four electoral college votes as he won the most votes in the 2nd Congressional District of Maine. In a similar vein, John McCain won the state of Nebraska at the 2008 Presidential Election, but Barack Obama took one of that state’s five electoral college votes as he won the 2nd Congressional District in that state in that election.
“Winner Takes All” apportionment of electoral college votes: In the electoral college system, the candidate who wins a state takes all the electoral college votes allocated to that state (with the notable exception of the Maine-Nebraska system), irrespective of the margin that they win that state by. On November 8th, Donald Trump won a number of key swing-states by a margin of around one percent, or even less, including Michigan, Wisconsin, Pennsylvania and Florida. If a small proportion of Trump voters in those states had voted instead for Clinton, or even opted not to turn out to vote, then Hillary Clinton could well have won those states and gone on to win the electoral college. That being said, Clinton did win the state of New Hampshire by a very small margin also (of just 2,736 votes, or 0.37% of the votes cast), while she also had a relatively narrow win in the state of Minnesota (a 44,765 vote winning margin, or 1.52% of the votes cast).
But the overall sense here is that the Trump vote was much more efficiently distributed, meaning that fewer Trump votes were “wasted” in the election. In other words, a larger percentage of the votes won by him in November 2016 were effective votes, which are defined by Johnston, Rossiter and Pattie (2005: 959) as “the votes that bring victory, calculated as the number of votes won by the second-placed candidate”. All other votes are either wasted votes – votes won by a candidate in a state that they lose and hence do not result in Electoral College votes – or surplus votes – votes that are “additional to the number of effective votes needed for victory” and hence do not directly result in Electoral College votes either. As it transpired, only 40.0% of all the votes cast at the November 2016 election were effective votes, with the other 60.0% being either wasted votes (Trump votes in states won by Clinton, Clinton votes in states won by Trump, votes for “Third Party” candidates) or surplus votes (Trump/Clinton surpluses in the states that they won). In the case of Trump, over 32.6 million of the votes won by him (51.8% of the total number of votes won by Trump) were effective votes – while 8.3 million Trump votes were surplus votes (13.3% of the total number of votes won by Trump) and 22.0 million Trump votes were wasted votes (34.9% of the total number of votes won by Trump). (This analysis controls for the fact that Trump won Maine Congressional District 2 and hence only a portion of the votes won by him in Maine were wasted votes, despite that state being won by Clinton.)
By contrast, a large proportion of Hillary Clinton votes were either wasted or surplus votes (i.e. were not used/needed to ensure that she won a state and all its electoral college votes) in terms of very big wins in large “blue states”, such as California (won by a margin of nearly 4.3 million votes) and New York (won by a margin of over 1.7 million votes). But a large number of Clinton votes were also wasted in terms of narrow losses in states such as Florida (where she won over 4.5 million votes, but these did not translate into any electoral college votes), Pennsylvania (where she won c.2.9 million votes), Michigan (where she won just under 2.3 million votes) and Texas (where she won just under 3.9 million votes). Just over 22.1 million of the votes won by Clinton (33.7% of the total number of votes won by Clinton) were effective votes – while 11.2 million Clinton votes were surplus votes (17.0% of the total number of votes won by Clinton) and 32.4 million Clinton votes were wasted votes (49.3% of the total number of votes won by Clinton). All of the votes won by “Third Party” candidates, such as Gary Johnston, Jill Stein and Even McMullin, would be classed as wasted votes, given that no Electoral College votes were won by “Third Party” candidates at the 2016 election.
Trump’s ability to ensure that there was a more efficient distribution of his vote meant that there was a resultant bias in the electoral college towards him. He was the more effective candidate in terms of winning votes in states where he/she needed to win votes. He managed to win a larger share of the electoral college votes than Clinton did, despite winning a notably smaller share of the national vote. Had his share of the national vote been exactly the same as Clinton’s (Clinton’s share of the national/popular vote was 2.1% higher than Trump’s share), the likelihood is that he probably would have also won in the states of New Hampshire and Minnesota (assuming a uniform swing from Clinton to Trump across all 52 states to equalise their vote shares), which would have added an extra 14 electoral college votes to his tally. This would have meant that Trump would have won 102 more electoral college votes than Clinton would have won (winning by a 320-218 margin) if both candidates had ended up on exactly the same share of the national/popular vote. Thus, in 2016, there was a 102-vote bias in the Electoral College favouring Trump, if he and Clinton had equal vote shares. (In the case of the 2012 election, the bias in the Electoral College system would have favoured the Democrats, as Obama would have won 32 more Electoral College votes than Romney would have if both candidates had won the same level/number of popular votes across the state.) By contrast, as discussed by Johnston, Rossiter and Pattie (2005), in 2000, there was a 52-vote bias in the Electoral College favouring George W. Bush, if he and Al Gore had equal vote shares, but there was a 40-seat bias in the Electoral College in 2004 that would have favoured John Kerry, relative to George W. Bush, if Kerry had managed to win the same number of popular votes as Bush in that election. (Bush did win the popular vote at the 2004 election, but lost to Gore in the popular vote at the 2000 election.)
The “winner takes all” system of awarding all the electoral college votes in a state to the candidate who wins that state plays a large role in accounting for such bias. So, what would have happened if each state’s electoral college votes had been apportioned proportionally on the basis of the percentage share of the vote won by each candidate – and assuming that certain thresholds were also applied (i.e. candidates only get electoral college votes if they win 5% of the national vote and/or 10% of the vote in an individual state)?
The story here is that Hillary Clinton would have won a narrow win in the electoral college by a 270-267 margin (Evan McMullin, an independent candidate, would have won one of the six electoral college votes in Utah).
Table 2(a): Number of electoral college votes that would have been won by each candidate if these had been apportioned proportionally (PrEV), as contrasted with the number won by each candidate in the actual election (“winner takes all”/”first past the post”).
As Table 2(a) shows, in a proportional allocation scenario the number of electoral college votes won by Clinton and Trump would have been evenly split in the key “swing states” that propelled Trump to victory on November 8th, namely Pennsylvania, Wisconsin and Michigan, as well as in Iowa. Trump would have won just one extra electoral college vote in Florida and North Carolina and two extra electoral college votes in Ohio. The gains made by Clinton in these states would have been offset by Trump gains in states that she won on November 8th – including, obviously, the closely contested states of New Hampshire and Minnesota, but also the very large “blue states” of California (where Trump would have taken 19 electoral college votes in a proportional allocation scenario) and New York (where Trump would have taken 11 electoral college votes).
What if both Clinton and Trump had won exactly the same number of popular votes/the same share of the national/popular vote? Assuming a uniform swing to Trump across all 50 states (and the District of Columbia) to bring his vote share into line with that of Clinton’s, this analysis was run again.
Table 2(b): Number of electoral college votes that would have been won by each candidate (in a scenario where both Trump and Clinton won the same share of the national vote) if these had been apportioned proportionally (PrEV), as contrasted with the number won by each candidate in the actual election (“winner takes all”/”first past the post”).
In this scenario, Trump would have enjoyed a very narrow win in the electoral college (Table 2(b)) – by a 270-267 margin over Clinton, with one of the electoral college votes in Utah again going to McMullin. This would have meant that Trump would have just about attained the necessary 270 votes required to win the electoral college, but this very narrow margin would have left him at the mercy of faithless electors and ultimately could have resulted in a scenario the final decision on who would be president and vice-president would have rested with the Congress and with the Senate.
Obviously a different scenario for allocating electoral college votes would have fundamentally changed the campaign strategies, or the geography of the election campaign. There would have been a greater focus on closely contested states with odd numbers of electoral college votes, given that a win in these states would have resulted in a candidate getting an advantage (albeit by one electoral college vote) over their opponent. But there would also have been a bigger play made for the very large states, such as California, Texas, Florida and New York, on the basis that a percentage gain/swing in your favour could well translate into a gain of a number of electoral college votes.
Differential Turnout Levels: Differences in voter turnout levels can also act to skew the relationship between the popular vote and the electoral college votes won by each candidates. If voter turnout levels are especially high in the states won by a certain candidate, this may act to partly explain why their opponent fares better than would be expected in the electoral college vote if based on their number of votes nationally/their share of the popular vote. Unfortunately, as of now, there are no final/definitive official voter turnout figures available for each of the 50 states to tease out whether turnout levels were higher, on average, in the states won by Hillary Clinton, which would then support the case that the bias in the electoral college towards Trump could be part explained by the geography of voter turnout (at the state-level). Data collected by the United States Elections Projects website does suggest that the highest voter turnout levels at the state-level for the 2016 contest were in a number of “swing states” that were won by Hilary Clinton, namely Minnesota, Maine, New Hampshire and Colorado.
Number of Electoral College Votes allocated to each State: The number of electoral college votes allocated to each state is based on that state’s number of Senators and the number of Members of the House of Congress allocated to that state. As the number of Congress members is roughly proportional to population – i.e. the largest state, California, has 53 members, the next largest state, Texas, has 36 members, the next largest states, Florida and New York have 27 members… – this means that there is a relationship between a state’s population and the number of electoral college votes that are allocated to it. But it is not a neat relationship and there is a bias towards the smallest states here.
The very small states, such as Wyoming (which accounts for just 0.2% of the USA population), are entitled to three electoral college votes, as each state will get two Senators and at least one member of the House of Congress. Does this bias towards the smallest states offer an advantage to one party, or one candidate, over another? Well, not especially so. Some of the largest states, California, New York and Illinois, are blue states, or states that tend to be won by the Democrats. However, Donald Trump won most of the other large states (including Texas, Florida, Pennsylvania, Michigan, Georgia and North Carolina) on November 8th. By contrast, Trump won in a number of the “red states” in the Prairies region, which have relatively small populations, such as Wyoming, North Dakota, South Dakota, Montana and Alaska. But, Clinton also won in a number of “blue states” in the North East, which also have relatively small populations, including Vermont and Delaware, as well as the District of Columbia.
Table 3: Number of electoral college votes that would have been won by each candidate if the number of electoral college votes per state was exactly proportional to that state’s population.
Table 3 shows that the manner in which electoral college votes are allocated to the different states did offer a slight advantage to Trump, but this only amounted to a minute advantage of 1.4 electoral college votes and this factor obviously did not dictate the final result in the electoral college to any significant degree.
“Third Party Candidates”: While third party candidates have not managed to win any electoral college votes at a US Presidential Election since the 1968 election (when George Wallace won 46 electoral college votes by means of winning the states of Georgia, Alabama, Mississippi, Arkansas and Louisiana), third party candidates have won different levels of support across the most recent presidential contests. The strongest performance by a third party candidate over recent decades came in 1992, when Ross Perot won 19.7 million votes, or 18.9% of the popular vote, even though he failed to win any states and any electoral college votes in that election. Ralph Nader, as the then official Green Party candidate, fared notably well at the 2000 contest (winning 2.9 million votes) and could have well cost Al Gore the presidency by taking potential Gore votes in the crucial close-states of Florida (Bush won by 537 votes – Nader won 97,488 votes) and New Hamsphire (Bush won by 7,211votes – Nader won 22,198 votes). However, prior to the 2016 contest, third party candidates did not fare well at most of the other elections during the 2000s; accounting for 2.6% of the vote in 2004, 1.5% in 2008 and 1.8% in 2012, as well as 3.8% of the vote in 2000, despite the strong Nader performance.
2016 marked a notably stronger performance for third party candidates – and most notably Gary Johnson of the Libertarian Party, who won nearly 4.45 million votes in this election (3.3% of the popular vote nationally), as well as Jill Stein of the Green Party, who won just over 1.45 million votes (1.0% of the popular vote). Evan McMullin only contested a number of states, but he won nearly a quarter of a million votes (21.3% of the votes) in the state of Utah. Admittedly, the pre-election polls had boded more favourably for these third party candidates. Some polls, taken a few weeks before the election, were suggesting that Johnson could win close to ten percent of the national vote and that Stein could win close to four percent of the national vote, while some polls also pointed towards a McMullin win in the state of Utah.
That being said, third party candidates won over 8.2 million votes (6.0% of the national vote) in this election, which shows that third party candidates had a much stronger impact in 2016 than in most of the preceding electoral contests. Did the presence of third party candidates skew the results, taking potential votes off one of the candidates from the two larger parties, particularly in key swing states? The jury is probably out on that. The general expectation was that Johnson might take votes off Trump, but that obviously did not happen to a sufficiently significant level in the actual election to skew the contest in the favour of Hillary Clinton. Indeed, a review of most opinion polls held before the election suggested that there was no significant advantage or disadvantage for either Clinton or Trump when the third party candidates were included in these polls. If there was a Johnston effect (and a McMullin effect), it may well have been to reduce the Trump vote but not to the point that it significantly impacted on Trump’s prospects in most of the key swing states, or close states, in this election, with the notable exceptions maybe of Minnesota and New Hampshire.
In terms of the actual election, there were a number of states (14, in all) where the presence of third party candidates could have theoretically shaped/skewed the final result – i.e. where the number of votes won by third party candidates was larger than the winning margin in that state. These states included Arizona, Colorado, Florida, Maine, Michigan, Minnesota, Nevada, New Hampshire, New Mexico, North Carolina, Pennsylvania, Utah, Virginia and Wisconsin. In some of these cases (e.g. Virginia, North Carolina and Arizona) the losing candidate would have had to have won nearly all of the third party candidate votes to turn the result around in their favour. But in the very close contests, there could be an argument to suggest that the winning/losing of that state could have been influenced by the potential number of votes that were lost by one of the main contenders to one, or more, third party candidates. But the cases in which Clinton could have been potentially handicapped by third party candidates (e.g. Michigan, Pennsylvania, Wisconsin and possibly also Florida) are offset, in part, by the number of cases where Trump likewise could have been potentially handicapped by third party candidates (e.g. New Hampshire, Minnesota, Maine and possibly also Colorado).
It is also worth noting that a strong vote for a third party candidate, or a number of third party candidates, in a specific states can have the effect of ensuring that the candidate who wins that state can do so with a much smaller vote/smaller share of the vote than they would have needed if the contest was effectively just between them and the main rival. In the aforementioned case of Utah, for instance, Mitt Romney won that state in 2012 with 740,600 votes, or 72.6% of the votes in that state. McMullin’s strong vote in Utah in 2016 ironically meant that Trump could still win that state with just 45.1% of the vote, or 515,231 votes. The strong McMullin vote meant that the margin by which the Republican candidate won the state of Utah fell from a 47.9% winning margin in 2012 to a winning margin of just 17.9% in 2016. But the reduced size of the Republican winning margin ultimately had no impact whatsoever in terms of which candidate took all of Utah’s 6 electoral votes. Effectively, the smaller Republican margin/surplus just meant that Trump was winning Utah while amassing a smaller number of “wasted votes” (surplus votes) in the process.
In terms of the overall national picture, it can be seen that third party candidates won, on average, 6.0% of the total number of votes cast – marking, as suggested already, a notable improvement on the performance of third party candidates in presidential election contests relative to those other contests held during the 2000s. Ironically, third party candidates actually, on average, fared better in the states won by Hilary Clinton (winning 6.7% of the vote in these) than in the states won by Donald Trump (winning 5.4% of the vote in these). These statistics would suggest that third party candidates did not play a significant role in accounting for the levels of disproportionality and bias evidenced at the 2016 election.
Conclusion: The geography of elections is a fascinating topic in itself. However, a case such as the 2016 USA Presidential Elections shows how a grasp of electoral geography can help towards understanding how Donald Trump won the election despite losing in the popular vote and indeed can help understanding how bias operates in all types of electoral systems. Having looked at different aspects of the US electoral system, it can be suggested that Trump’s win in the electoral college could potentially have been shaped by the (somewhat) disproportional manner in which electoral college votes are allocated to the larger/smaller states, as well as the impact of differential turnout levels and the impact of third party candidates. But the main conclusion here is that the “winner takes all” nature of the US electoral system was the main factor at play here – the fact that states award their electoral college votes on a “winner takes all” basis offered a significant advantage to Donald Trump in this election, especially given his ability to pull off narrow wins in highly competitive states, such as Michigan, Wisconsin, Pennsylvania and Florida. Had each state’s electoral college votes been instead awarded on a proportional basis, then Hillary Clinton would have enjoyed a narrow 271-266 win in the electoral college, with one electoral college vote (in Utah) been won by Evan McMullin. But, of course, it could be argued that, in this scenario, the geography of the Trump campaign (and the Clinton campaign) would have been significantly altered to potentially reshape the final result more so in their favour.
Suggested Further Reading/Useful Resources:
- Johnston, R.J., Rossiter, D.J. and Pattie, C.J. (2006B). Changing the scale and changing the result: Evaluating the impact of an electoral reform on the 2000 and 2004 US Presidential elections. Political Geography, 25(5), 557-569.
- Johnston, R.J., Rossiter, D.J. and Pattie, C.J. (2006A). Disproportionality and bias in the results of the 2005 general election in Great Britain: evaluating the electoral system’s impact. Journal of Elections, Public Opinion and Parties, 16(1), 37-54.
- Johnston, R.J., Rossiter, D.J. and Pattie, C.J. (2005). Disproportionality and bias in US Presidential elections: how geography helped Bush defeat Gore but couldn’t help Kerry beat Bush. Political Geography, 24(8), 952-968.
- Dave Leip’s Atlas of US Presidential Elections website; http://uselectionatlas.org/