Saturday, August 20, 2016

The United States Electoral College and third party candidates

As most of my readers will know, elections to the Presidency of the United States, as well as approximately a third of the Senate, the entire House of Representatives, and various other state and local offices will be up for election on November 8 (by tradition, the first Tuesday in November).

While primaries for some lower-level offices are yet to take place, Presidential candidates for both major parties have been chosen through a series of primaries and caucuses. The Democrats have chosen Hillary Clinton, the early favourite, though she faced a stiff challenge from left-wing Bernie Sanders. On the other hand, a crowded Republican field narrowed to one candidate; controversial businessman Donald Trump, who had little support from the party leadership, won the nomination.

Trump's views represent a substantial diversion from traditional Republican thinking on a variety of issues. The most notable one of these is free trade; Mr Trump is opposed to it, but Republicans have traditionally supported free trade. Trump's tax plan is also a bone of controversy; while it would reduce tax rates, normally something the Republicans would support, it also does not make cuts to government programs like Medicare (health insurance for the elderly) and Social Security (the government retirement program), thus meaning that the deficit would increase.

These two issues have led a number of Republicans to look elsewhere in the presidential field. While some will vote for Ms Clinton, one other candidate gaining traction is Gary Johnson, of the Libertarian Party. The Libertarians have normally been a very minor political presence in the United States; occasionally winning a seat in state legislatures. The party's best performances have been in Senate and House elections when support for either the Republicans or Democrats in a state or district is so overwhelming that only one of the major parties puts up a candidate; when a Libertarian runs in one of these seats, they normally win a substantial share of the vote.

This year, Mr Johnson is currently polling at around 6-8% in national polls. Mr Johnson receiving support from dissatisfied Republicans is no doubt helped by his credentials as a Republican; he served as the Republican governor of New Mexico, and his running mate William Weld served as the governor of Massachusetts, also as a Republican. While this does not look like much, it would represent the strongest performance ever, by a long way, for a Libertarian in a presidential election.

Some Democrats are also not wholly happy with Hillary Clinton. Some supporters of Sanders, a self-declared democratic socialist, view Ms Clinton as too right-wing, and are also looking elsewhere. This has caused a much more muted increase in support for Green Party candidate Jill Stein. Not all polls include Ms Stein, but those that do mostly show her receiving 2-4% support; a low figure, but a substantial increase over the 0.36% she received in 2012 (though she was not on the ballot in all states at this election).

Minor candidates and Presidential elections

In countries where a President is elected, there are two main ways this is normally done; the plurality system, or the two-round system.

Under the plurality system, voters vote for one candidate, and the candidate with the most candidates is elected. Such a system is used in Taiwan, South Korea, and Venezuela, amongst others. In a close nationwide race, this system has a tendency to lead to two main contenders, as supporters of any weaker third candidate would have strong tactical incentives to vote for the member of the top two they most prefer.

However, when voters do not do this, the system can lead to candidates from one political faction being elected despite a majority of the electorate not supporting that faction. For example, at the Taiwanese presidential election in 2000, the pro-Taiwanese independence Democratic Progressive Party nominated Chen Shui-bian, who won 39.3% of the vote. However, two candidates from the opposing bloc (generally in favour of Chinese reunification) ran; the Nationalist candidate Lien Chan won 23.1%, and independent James Soong won 36.8%. For this reason, despite an apparent majority of voters supporting a pro-unification President, the pro-independence candidate won.

Most countries that have an elected President use a runoff system. Under such a system, voters vote for one candidate in the first round. If one candidate receives a majority, that candidate is elected; if not, a second round takes place, normally a couple of weeks after the first, when voters choose between the top two (in most cases) candidates.

This means that there is no incentive for tactically abandoning your most preferred candidate for the lesser of two evils in the first round, as unless a candidate in the first round wins 50%+1 (in which case choosing the lesser of two evils would not have mattered) you will be able to make that choice in the second round.

However, such a system can have its disadvantages. For example, at the 2002 French presidential election, the centre-right Union for a Popular Movement candidate, incumbent President Jacques Chirac won 20% of the vote. It was expected that he would face Socialist candidate Lionel Jospin in the second round, but a large number of left-leaning voters voted for other leftists in the first round; for example, the Workers' Struggle candidate won 5.7%, the Green won 5.2%, and the Communist won 3.4%. This meant that Jospin received only 16.2%, behind far-right candidate Jean Le Pen, who won 16.9%.

In the runoff, leftists voted for Chirac over Le Pen, resulting in Chirac winning 82% of the vote. These results demonstrate that there are still incentives for tactical voting under a two-round system; a voter may tactically vote to make the opponent of their most preferred candidate in the runoff an unpalatable radical.

The United States has a somewhat different, and currently unique, manner of electing the President. Each state is allocated a number of 'electors' equivalent to the number of members that state has in the House of Representatives and Senate combined. Each state has two Senators, regardless of size, but the number of members each state has in the House is proportional to its population. This method of allocation over-represents smaller States; for example, the state of Montana has 341,333 people per elector, but the state of California has 705,454 people per elector.

This means that on Election Day, voters are not voting for a presidential candidate, but rather electors who pledge to vote for that presidential candidate. There have been cases in the past where these electors have not voted for the candidate they were listed on the ballot for, though this has never changed the outcome of a presidential election.

Potential incentives of the Electoral College

So, what incentives would the Electoral College have, in terms of small party candidates? Well, over time, certain parties have become wholly dominant in states. For example, the Democratic Party has won the state of Massachusetts, in nearly all cases with a double-digit margin of victory, in all elections since 1984. The same goes for the Republican Party in Texas. This is why presidential elections are focused on swing states; those states where the race between the top two candidates is especially close, like Pennsylvania and Florida. Winning extra votes in a safe state is irrelevant, as the number of electoral votes (which are what matter) do not change. 

If you live in a safe state, even if there is a close vote nationwide, you have no incentive to tactically vote to support the 'lesser of two evils', given that the winner of your state is virtually determined in advance. In theory, then, people in safe states should be more willing to vote for third party candidates, given that there is little danger of them 'wasting their vote' and electing someone they dislike.

But does this work in practice? Let's look at two third-party candidates who were both totally unlikely to win; John Anderson in 1980 and Ross Perot in 1996. Mr Anderson was a moderate Republican congressman opposed to the right-wing views of party nominee Ronald Reagan, while Mr Perot was a populist businessman running on a platform of deficit reduction. Perot had run in 1992, but this campaign may not have had the same tactical messaging for the voters, since Perot was in first place in some early polls.

The below charts show a comparison between the vote shares for these candidates by state, compared to the margin between the top two candidates, which represents how close a state was.

There is something of a correlation between the two factors, with the correlation being stronger with Perot than with Anderson (the precise correlation factor being 0.13 for Anderson and 0.30 for Perot). However, there are many other factors which vary these numbers, and it is still quite a weak correlation, which makes it somewhat difficult to draw any conclusions.

It is interesting, though, to consider the states that the third party candidates will do best in. Mr Johnson has public profile in New Mexico, which, though it has been quite close in the past, was comfortably in the Democrat column in 2012 (a margin of ten points to Obama). The state's high Hispanic population is also unlikely to approve of Mr Trump's rhetoric on immigration, which would mean that both campaigns would be sensible to ignore the state.

Mr Johnson's running mate, Mr Weld, is the former governor of Massachusetts, a very safe Democratic state (despite the Republican candidate for President coming from the state, it went Democratic by 23 points). While these two cannot choose their home states, for the Libertarians their states make them sensible nominees, as their home state boosts will not be cancelled out by tactical voting.

The lower statewide profile of the Green ticket (Ms Stein has run for Governor of her home state of Massachusetts, but with little success, and her highest office remains being in the Lexington town meeting, while her running mate Ajamu Baraka does not appear to have run for any office) makes this factor less relevant for them. Nonetheless, it will help Ms Stein somewhat that the state where she has the most profile (relatively speaking) is a safe state, where she will have somewhat clearer air to campaign.

Saturday, July 30, 2016

Australian election, 2016-Xenophon, Hanson and Singh

Australia held a federal election on July 2 of this year for all members of the Federal Parliament. The Australian Federal Parliament consists of a 150-member House of Representatives and a 76-member Senate. The government is responsible to the House of Representatives, which is elected through preferential voting in single-member districts for a three year term (though it may be dissolved at any time by the Governor-General, acting on the advice of the Prime Minister).

However, all legislation must pass through the Senate, which is elected through the single transferable vote, though with substantial modifications described at the end of the article. Elections to the Senate are usually for 40 seats at a time (six Senators per state for six-year terms, plus four Territory senators elected for three-year terms), but if the Senate rejects a bill passed by the House of Representatives twice, the Prime Minister may advise the Governor-General to call an election for all members of both houses.

The result was the narrow return of the incumbent Liberal/National coalition government, led by Prime Minister Malcolm Turnbull. In the 150-member House of Representatives, the party will definitely have 76 seats, which is a majority of one. If the Speaker of the House is a Liberal, which seems likely, the government will have a majority of 75-74.

The seat of Herbert, in the north of Queensland, is the only seat that could still likely switch between the two parties. Labor candidate Cathy O'Toole narrowly edged out Liberal incumbent Ewen Jones by eight votes on initial counting; the margin increased to 35 following a recount. While a narrow margin of this sort could theoretically be upheld, any minor error in counting or voting that would be ignored in a seat with a wider margin could lead to the election heading to the courts.

Even if Herbert is won by the government, the 77 votes in the House will represent a knife-edge majority almost as close as that won by Robert Menzies in 1961 (62 Liberal/Country, 60 Labor with full voting rights). The Turnbull government will have the advantage, though, that the Labor Party do not have all the remaining MPs. The House of Representatives crossbench consists of one Green, one Katter's Australian Party (Bob himself), two Independents (Andrew Wilkie in Denison and Cathy McGowan in Indi) and one member of the Nick Xenophon Team.

Whether these members will support the government consistently is unclear. The AEC conducts preference counts between the two major nationwide parties in each seat for statistical purposes. At the last election, McGowan's division of Indi split 59-41 to the Liberals, and Katter's division of Kennedy went 67-33. The Xenophon Team's seat of Mayo split 63-37 to the Liberals, though the Xenophon Team's preferences are more of an unknown quantity. Consistent opposition to the government, or supporting the installation of a Labor government, would likely be harmful to the future electoral prospects of these MPs.

The Xenophon Team's result

Before this election, I wrote about the Xenophon Team's prospects. I said that the difficulty Xenophon had in transferring his support to his chosen candidates, demonstrated in the 2014 South Australian election, would make a Xenophon lower house seat unlikely. Nonetheless, the party appears to have won a seat. So, how big was the difference between Xenophon Team House candidates and the support for the man himself?

Measuring the dropoff between Xenophon's House and Senate support simply by comparing their share of the votes in both houses is the obvious way to do this; however, it is something of an unfair comparison, given that there were an average of 6.5 candidates per House seat compared to 23 groups in the Senate. Nonetheless, here it is, though this is based on somewhat incomplete vote totals (as counting and data entry still continues).


In only three seats (Mayo, Barker and Grey) did the Xenophon House candidate outperform the Senate ticket, even with increased choice on the Senate ballot paper. Incidentally, these three seats were the strongest for Xenophon Team House support. On average, the Xenophon Team performed 1.24% stronger in the Senate than the House. This suggests that the dropoff between Xenophon House and Senate support is smaller than the 2014 SA election would suggest, but still substantial.

Incidentally, if you calculate Xenophon Team support as a percentage of the total for the parties that contested all the House seats (that is; Labor, the Greens, NXT, the Liberals, and Family First) in both houses, you get a somewhat different result.

This result moves Grey into the seats where the NXT Senate ticket performed better than the House, and increases the average to 2.3%. In general, NXT did not make much of an impact in the other states, winning an average of 1.8%, and probably not enough to win a seat (though a very strong performance on preferences might give them a chance in Western Australia). However, they won 21.6% in SA, giving them 2.8 quotas and likely three seats.

One Nation, under Palmer?

The other substantial new minor party presence in this election was the One Nation party. This party was formed by Pauline Hanson, elected as an independent in the seat of Oxley in 1996 (with a nearly 20% swing towards her) after being disendorsed by the Liberal Party for making controversial comments about indigenous Australians. One year later, she formed a political party, called One Nation. This party espoused anti-immigration views, though, as is not uncommon with this sort of party, supported somewhat economically leftist policies, especially opposition to privatisation and free trade (a lot of her views are laid out in her maiden speech to the House).

This party first contested the 1998 Queensland election, winning eleven seats and 22.7% of the vote. However, the party soon split into two, with some MPs joining the City Country Alliance (warning: vile web design), and the party lost most of its seats at the 2001 election. 

At the federal election of 1998, the party won 8.4% of the House vote (14.3% in Queensland) and 9% of the Senate vote, though weak preference flows meant that this only translated into one seat, in Queensland. Hanson ran in the seat of Blair, and won 36% of the primary vote to 25.3% for Labor, 21.7% for the Liberals, and 10.2% for the Nationals; however, the Liberal candidate received enough preferences from the National to pull ahead of the Labor candidate, whose preferences elected the Liberal over Hanson 53-47.

This election represented a peak for One Nation, whose vote roughly halved at the next election, and declined to below 1% after that. Hanson continued to contest elections, though often not as a member of the One Nation party; she ran as a member of her own United Australia Party for the Senate in 2007, winning 4.2%. She came close to winning a seat in the New South Wales upper house in 2011, ran for the Senate in 2013 (and would have potentially won had voters not confused the Liberal Party with David Leyonhjelm's Liberal Democrats), and came within 0.2% of winning the Queensland state electorate of Lockyer in 2015).

But now, she appears to have certainly returned to the Senate, as One Nation has secured 9.1% of the Senate vote in Queensland, comfortably above the quota for election to one of the twelve seats. One Nation are also in contention to win a seat in New South Wales and Western Australia with about 4%. They also missed out on a Tasmanian Senate seat by 140 votes off 2.57% of the primary vote.

So, where does this One Nation vote come from? Well, the below chart shows how the One Nation vote correlates to the vote of the party whose vote dropped most dramatically at this election: the Palmer United Party. This party was formed by mining billionaire Clive Palmer leading up to the 2013 election, and won two Senate seats (with an extra member elected in the Western Australia re-election) and one House seat (Palmer himself won the seat of Fairfax). However, two of the party's Senators left, and Palmer did not recontest Fairfax. None of the party's Senate tickets secured as much as half a percent of the vote.

The below chart shows the correlation between the Senate vote for the Palmer United Party in 2013, and the Senate vote for One Nation (based off current counting) for this year.

The two variables have a correlation coefficient of 0.73, suggesting a relatively close correlation. However, this likely reflects the fact that in most electorates, both One Nation and the Palmer United Party are/were a relatively minor presence. In inner-city electorates, both One Nation and the Palmer United Party having roughly 1% of the vote is not an especially meaningful correlation. The below chart shows the correlation for the divisions where the Palmer United Party received more than 7% of the Senate vote (the vast majority of which are in Queensland).
Here, we see much greater variation, and a much weaker correlation coefficient of 0.22. While this is dragged down somewhat by the presence of Jacquie Lambie in Tasmania (presumably taking some of One Nation's constituency), there are only two Tasmanian electorates in that dataset. While both the Palmer United Party and One Nation have substantial support in Queensland, they do appear to have substantial differences in where they win support there.

Lisa Singh wins on BTL votes-could it happen elsewhere?

So far, only two jurisdictions have declared Senate results; Tasmania, and the Northern Territory. The Senate electoral system has changed this year. Amongst other changes, voters now only have to number six boxes in order to cast a formal vote below the line (for individual candidates), though the ballot paper says twelve.

When tickets for the Senate were drawn up for Tasmania, two controversial decisions were made. Labour incumbent Senator Lisa Singh was demoted to fifth on the ticket, a very difficult position to win, while Liberal Tourism Minister Richard Colbeck was moved to the same placing. 

As casting a vote for these candidates individually is much easier under the new system, both Colbeck and Singh either ran, or had campaigns run for them, to encourage people to vote for them below the line. In the end, Singh was returned to the Senate, with 6.12% of the below-the-line vote, while Colbeck was not returned, despite winning 4% of below-the-line votes. The result represents the first time since the 1950s, and the first time since any type of above-the-line voting existed, that a Senator has been elected out of party ordering.

So, could the new system allow this to happen in other parts of the country? My answer is probably not. As I wrote when I was discussing the initial version of Senate reform, that would have required voters to number every box below the line, the states of South Australia and New South Wales both fairly briefly used the single transferable vote for multiple seats to elect their upper houses, without any form of above-the-line voting. While New South Wales required voters to number seventeen boxes, while all were required in South Australia, the systems shared one common characteristic; it was no harder to vote against the party ticket than to fully accept it. In both of these cases, the vast majority of voters appear to have accepted the party ticket.

The fact is that Tasmania is a much smaller state than any on the mainland, and running a statewide campaign for below-the-line votes is therefore much less costly than in any of the mainland states. Given that political parties have expressed an unwillingness to encourage such campaigning, doing it on a larger scale would prove difficult. Colbeck and Singh had the wind to their backs in another way; this election is a double dissolution, for twelve Senators, while most Senate elections would be for six, making it even harder to win a seat.

Though the equal distribution of votes between a party's candidates is often advantageous (if too many votes are concentrated on one of a party's candidates, it can lead to the other candidates of that party being excluded early, thus costing that party seats), Australian political parties appear to remain opposed to individual campaigns. The difficulty a Senate candidate's campaign would face in getting attention in the midst of a House election, with the Prime Ministership at stake, would add to these difficulties. Outside of Tasmania, it will be unlikely for any Senate candidates to be elected outside their party's ranking in the near future.