It looks like the curtain has finally come down on Australia’s cliffhanger election, with Malcolm Turnbull’s Liberal/National coalition hanging on by way of deals with several one-seat parties.
One underreported aspect of the results, though, is the bias in them.
That bias favours the right-wing parties, and occurs in basically all single-member-district electoral systems, including Australia’s Alternative Vote system and New Zealand’s old First Past the Post system.
The reason for the bias is that there tend to be tighter concentrations of lefties in inner city and low-income electorates, compared to the more even distribution of lefties and righties in the suburbs. That is: fewer right-leaning voters choose to live in places like Mangere, compared to the number of lefties who choose to live in places like Ilam.
As a result left-leaning parties tend – on average – to win their seats by a greater margin, compared to right-wing parties, meaning left parties win more votes per seat. That’s the bias.
There’s plenty of evidence in support of this bias in New Zealand history. Across the twenty elections we held under FPP from 1935-1993, Labour averaged 14,230 votes per parliamentary seat, whereas National averaged only 13,770 votes per seat. That difference, of almost 500 votes per seat, is the bias.
The place this bias shows up most obviously when the popular vote goes one way, but the election goes the other way. That’s what looks to have happened in Australia. This morning, Labor was ahead 50.1% to 49.9%, this afternoon, the coalition re-took the lead 50.3% to 49.7%. Whatever the final result comes out as, it will be incredibly close.
By rights, the seats should also be tie, yes?
But the seats aren’t really close at all. The Coalition is solidly ahead, looking to have 77 seats, while Labor has 68 and the remaining few are scattered among the smaller parties.
This has happened before. In Australia, Labour has been on the wrong end of winning the popular vote but losing the election on five separate occasions – 1940, 1954, 1961, 1969, 1998. Only once has the reverse happened.
When New Zealand had FPP elections, the same thing happened twice, in 1978 and 1981. Both times it cost the Labour party and helped the National party. In the US it famously happened to Al Gore in 2000. It happened to Congressional Democrats in 2012. And so on.
This is one of those patterns that’s reasonably well known in the political science community, but almost entirely unknown outside it.
Jowei Chen, my former colleague at Michigan, and his coauthor Jonathan Rodden, showed this effect really convincingly in a lab environment, using a set of computer simulations. The computer drew lots of electorate boundaries in Florida, using non-partisan criteria of being contiguous and small. Then it used real vote data to run trial-heat elections using those boundaries. The result: The Republican seat share usually outperformed its vote share, while the Democrats suffered the flipside of that bias.
Should Australians be outraged that the seat split didn’t follow the vote split?
If you win the popular vote in a democratic election, shouldn’t you normally expect to win the election as well, whether you're a leftie or a rightie? And if you tie the popular vote, shouldn;t you also tie the seat split?
Is there any good, democratic reason not to expect this?