How people elect parliaments
The very close result in last night’s special election for the vacant 18th congressional district of Pennsylvania looks set to attract a lot of attention to the idea of ranked choice voting.
In a major snub to President Trump and the Republican-controlled Congress, a seat with a nominal 20% lean towards the Republicans appears now to have been seized by Democratic candidate Conor Lamb.
At close of counting – with 2 precincts out of 593 still to report and the remaining absentee ballots being progressively counted – Democratic candidate Lamb is 850 votes ahead of his Republican rival Rick Saccone, with 111,875 votes to 111,028. [*Update at end of this post.]
But there was a third candidate, Drew Miller of the Libertarian party. Miller won 1,351 votes. Since Libertarian voters are generally seen as somewhat right-of-center, the case is sure to get made that had Miller not run, or alternatively had the electoral system used ranked choice voting, the Republicans might just have saved the seat.
With ranked choice, or preferential voting, voters indicate additional next-preferences, marking the candidates 1, 2,3 and so on. Using such a system, the ballots of Millers 1,351 voters could have been scrutinized again and transferred to either Lamb or Saccone.
All elections in Australia use ‘preferential voting’ – as it is known there.
Australia is the world’s preeminent user of the technique, applied both to single-member division elections and in a multi-member division form, known as the single transferable vote (STV). The voting method has been used for almost a century.
Preferential voting allows minor parties and independents to run without fear of ‘strategic voting’ in single-member division systems, which are used in most of the lower houses of Australia’s parliamentary systems of government at the national level and in the Australian eight states and territories.
Preferential voting gives voters a much fairer range of free choice, but if the minor candidates do not poll well, the preferences marked on each ballot will ensure that the final choice between the two major party candidates reflects overall voter opinion.
In its multi-member STV form, ranked choice election results are even more representative, distributing seats in the parliaments in fair proportion to large and small parties and also to independents.
Would ranked choice voting on its own have changed last night’s Pennsylvanian election result? Probably not. On these numbers, Saccone would have needed to win around 81% of ‘Libertarian preferences’ (as Australians would term the ballots) – 1,101 out of 1,351 – to overtake Lamb.
If voters did not actually mark any second preference between the Republican and Democratic candidates, the required percentage would be even higher, since Saccone would need to make an absolute gain of 851 votes.
Most Australian voting systems require voters to mark preferences for every candidate, but the fairer form of voter choice is ‘voluntary preferencing’, where voters can leave some candidates’ names blank if they wish. (In Australia such ballots are termed ‘exhausted’.)
It is hard to find examples of US Congressional elections that would illustrate the effect of ranked choice voting. A race is not in issue if there are only two candidates from the major parties – no matter how close the result.
Minor parties and independents basically play very small roles in the US political system, compared to almost every other democratic country.
The US has Libertarian and Green parties, which typically score around 1% support. New York has a Conservative party and a Working Families party, which register votes, although they typically actually endorse Republican or Democratic candidates as if they were their own candidate. A Constitution party occasionally registers very small votes, and a Reform party made some headway about a decade ago, but has since abated.
Independent candidates occasionally get support in local communities, and indeed two US states currently have independent Senators: Maine (Senator Angus King) and Vermont (Senator Bernie Sanders).
But the dominance of the two major US parties suppresses the realistic option for voters to experiment with candidate choices other than from the two dominant parties.
To influence a single-member district race, the third party votes must obviously exceed in number the gap between the two major parties. Amazingly, in the three general House elections this decade – 2012, 2014 and 2016, covering 1,305 district elections – that has happened on just 39 occasions. There have been more occasions than that when an ultra-safe district has seen only one candidate nominate.
But seeing more third-party votes than the Republican-Democrat margin is far from enough to really influence the result of the election. Many of those instances involve a conservative or a Libertarian doing well in a race where a Republican is in the lead, so the ballot preferences flowing from the Libertarian tally would normally only increase the Republican’s lead. The same would occur where votes for a Green candidate would increase a lead held by a Democratic candidate.
There are really just a tiny handful of illustrative cases. In 2012, in the contest for the 1st district of Arizona, covering large non-urban areas of the state, the Democratic candidate Ann Kirkpatrick led the Republican candidate by 122,774 votes to 113,594 – a margin of 9,180 votes. A Libertarian candidate had another 15,227 votes. Under plurality rules, Kirkpatrick won the seat, which had been left vacant by a retiring Republican incumbent. Her rival Republican would have needed just over 80% of these Libertarian preferences to win the election (more if voter preferences exhausted to any extent). In rural Arizona, where Libertarians are probably fairly conservative, those numbers would probably close the gap considerably, but Kirkpartick would most likely have stayed ahead during the preference count and still won the seat.
In the same 2012 election, in the 6th district of Massachusetts, north-east of Boston, the Democratic candidate John Tierney led his Republican rival by 4,330 votes, while a Libertarian candidate won 16,739. The Republican would here have needed 63% of the preferences. That’s a bit more likely, except that in Massachusetts Libertarian voters are probably less strongly conservative overall. In any case, this one would likely have ended up fairly close.
In the 1st district of New Hampshire, which covers the south-eastern part of the state, in the 2016 election the Democratic candidate Carol Shea-Porter finished ahead of the incumbent Republican congressman Frank Guinta by 4,904 votes. Here there was also a Libertarian candidate winning 5,507 votes. But there was also an independent candidate – Shawn O’Connor – who won a massive 34,735 votes, and finally a second independent Brendan Kelly who won 6,074 votes. This is a much harder election to attempt to re-cast using imaginary preferential voting. It depends largely on how conservative the voters supporting the independents were.
Finally, there was the contest in 2012 for the 13th district of Illinois, covering rural areas in the south-west of the state. This election was actually won by the Republican candidate Rodney Davis over Democratic candidate David Gill by 1,002 votes. But also in the race was independent medical tech company financial officer John Hartman, who won a massive 21,319 votes. Gill’s target of preferences from Hartman voters would have been only just over 52% of them (before the impact of exhausted ballots). Given that Davis went on to win the 2014 election (with only two candidates) with 58% of the vote, it is fairly clear that the underlying voter sentiment of the district is pro-Republican, and thus that the hypothetical 2012 ‘Hartman preferences’ would probably have trended Republican, not Democratic.
The above are the most plausible four examples of House of Representatives November election races held since 2012 – from a total of of 1,305 contests – that show any real possibility of preferencing changing the actual election result. Even then, as we have seen, none of them give rise to a strong likelihood of a changed result.
But this tiny number of illustrations does not paint a true picture. The above examples are simply based on the real election data, because of course there is no other data. But had the US congressional electoral system involved voters marking preferences, it is completely certain that many more voters would have marked their ‘first preferences’ for minor parties and independents, and in turn minor parties and independents would have nominated more candidates, developed stronger political grass roots and performance over the years, and occasionally even won some seats.
In Australia in modern times around 8% of single-member division election results are determined by the counting of preferences; that is, that candidates who placed first on the first preferences did not go on to be the ones who won the final count of votes after preferences. This is an entirely normal and accepted part of Australian elections, and reflects the fair empowerment of voters who primarily support minor parties such as the Greens, populist movements and minor Christian movements, as well as independents.
Another result revealed by this analysis, and amply demonstrated by Australian elections, is that so long as an electoral system is based on single-member district, even using ranked choice voting does not guarantee fair representation for the supporters of minor parties. Nor does it assure the electorate of a legislature with a seat balance between major parties that accurately reflects the national sentiment when elections are close.
In Canada in 2016, newly elected Prime Minister Justin Trudeau privately wanted to change the electoral system by which national Canadian MPs are elected from plurality voting to preferential voting in single-member electorates, termed the ‘alternative vote’ (AV) in Canada and Britain. Since Trudeau’s Liberal Party occupies a position towards the center of the political spectrum in a three-party system, this voting system would have advantaged his party greatly – and unfairly – at future elections.
But Trudeau had gone to the 2015 election openly promising a reform process aimed at a achieving more proportional representation, with the choice of system to be worked out by public consultation during his first term. A year later, when virtually no-one else involved in the reform process supported his preferred voting method, the Prime Minister simply abandoned the entire reform exercise.
By contrast in the United States electoral reform advocates interested in ranked choice voting, such as the FairVote organization operating nationally and in several states, are much more interested in electing people to their legislatures through multi-member, not single-member, districts.
The Fair Representation Bill presented to Congress by Virginian Congressman Don Beyer would introduce exactly that change on a national level.
If the unsuccessful campaign of candidate Drew Miller in yesterday’s Pennsylvanian election has any lasting outcome, it might well be to convince many Americans – including all the Libertarian, Green and other voters – of the beneficial possibilities of Beyer’s bill, and other similar reforms.
[Update: with the final two precincts in the Pennsylvania 18th district special election now reporting, the result stands at a 579-vote win by Democratic candidate Conor Lamb. But more absentee votes are still to be counted.
According to Politico website, “the current vote count includes more than 3,000 absentee ballots from Allegheny County, the most Democratic-leaning of the four counties that make up the district, and 1,808 from Westmoreland County, a Republican-leaning county. Lamb carried Allegheny County, 57 percent to 42 percent”.
On those revised numbers, Saccone would have needed 73% of Libertarian preferences to overtake Lamb’s vote tally, down from 81%.]