When 50.1% isn't enough

I somehow managed to miss this very interesting piece by Simon Jackman in the Bulletin, showing that there’s a systematic bias in the Australian electoral system against Labor, such that they typically need 51-52% of the two-party vote to win government.

Simon only slightly alludes to this, but naturally, one can’t help wondering whether this is something that the Australian Electoral Commission should be trying to rectify when it does its redistributions. Or is there something natural about Australian political geography that means Labor will always have more votes locked up in its safe seats? 

This entry was posted in Australian Politics. Bookmark the permalink.

11 Responses to When 50.1% isn't enough

  1. Graham Young says:

    When you look at state electoral boundaries the bias appears to be in the other direction. From this you can either infer that redistribution commissions tend to favour governments, or that changes in demographic allegiances can lead to more or less chance biases in the system.

    If you look at Queensland and the parties as individual units you can see support for your thesis that it has to do with votes being locked-up in safe seats. The Liberal Party has approximately 10% of the seats in parliament, despite scoring around a quarter of votes. The Nationals don’t fare so badly. They have 19% on a similar percentage of the total. The Nats do better than the Liberals because their vote is geographically concentrated, whereas the Libs are spread out.

    But, when the tide eventually turns (unlikely as this may now seem) the Liberals may end up with a disproportionate number of seats compared to Labor because for the tide to turn key urban demographics that now make Brisbane wall-to-wall Labor (with the exception of Moggill and Clayfield) will switch.

    After thinking about this for decades I’ve come to the conclusion that with single member constituencies, no redistribution is ever going to reward 50.1% of the vote with government every time, and that trying to redistribute to make this a result is probably more prone to error than working with what you’ve got and taking account of geographical boundaries and communities of interest.

  2. Sinclair Davidson says:

    That is a very good piece – it deserves wide exposure. I really do miss the AFR’s “Lies and Statistics” column.

    I’m wondering is house prices play a role. As housing affordability declines (asuming that it is, and without indicating who’s to blame, IMHO: state government) so lower-income voters who are more likely to vote ALP are less likely to move house and become geographically dispersed leading to a concentrated vote. The way to test that would be to look at those constituencies with new housing estates work out a weighted flow of voters into and out of the constituency and compare voting patterns across polling booths. The AEC might have the data to do this, but I suspect it’s not in the public domain.

  3. Patrick says:

    Or, to use lefty-style discrimination against them – the better educated and more mobile you are the more likely you are to vote Liberal.

    A junk theory, but some form of the inverse could pass as received wisdom on eg Larva Rodeo.

  4. Shane Easson says:

    On a uniform swing of 4.8% the ALP requires 52.1% 2PP to win 76 seats, a bare majority in the HofR.

    The Commonwealth Electoral Act specifies that the Commissioners –after meeting the numerical requirements of the Act (that Divisions be within 10% of the quota at the commencement of a redistribution and within 3.5% of the predicted average quota in the State 3 yrs and six months after the quota is struck) may only give due consideration of community of interest provisions contained in the Act. The Commissioners thereby must not consider political outcomes. I think that’s how it should be and that it would be a mistake to take the South Australian route.

    NB The numerical requirements referred to above were measures passed by the Hawke Gvt in 1984. These effectively removed malaportionment —in the ‘one vote one value sense’.

    So how to account for the current bias? I put it down to two things: sitting member incumbency advantage and also the nature of the support given to both sides of politics at the last election.

    In the last decade or so sitting MP’s in urban as well as regional and country seats have been able to build a personal support base. The level of personal support varies according to the ability of the MP, the stability of the seat, the importance of local issues and so forth. However, I’d argue that the general political climate also influences the ability of the local MP to build a personal vote. If for whatever reason there’s a general mood that ‘times up’ as appears to be the case currently then this will affect the ability of an MP to sustain a personal vote. I think we saw evidence of this at the 2007 NSW State election particularly in the Sutherland Shire where the sitting MP’s for Miranda and Menai sustained swings against of up to 8%.

    To illustrate the second point on the nature of support let me refer to QLD and the 2004 Federal Election. Coming into the 2004 election the ALP held only 7 of the then 28 QLD Divisions. However, there were six QLD Coalition seats held by a margin of 4% or less and 10 seats held by 7% or less. Post that election, in which the ALP lost the seat of Bonner and there was a 2.2% swing against the ALP, ( taking its 2PP vote from 45.1% to just 42.9%, comfortably the lowest ALP 2PP of any State ), the ALP held only 6 QLD Divisions. More significantly, only one Coalition seat, the just lost by the ALP seat of Bonner was held by a margin of less than 4% and only a total of 4 seats were held by margins of 7% or less.

    Following the 2006 Federal redistribution for QLD the ALP’s position improved slightly. Whilst QLD gained an extra seat ( Flynn with a notional National Party majority of 7.8%) there are now two seats Bonner and Moreton held by margins of less than 4% and a total of 5 Divisions held by margins of less than 7%.

    Why? The swings in Coalition held marginal were greater than the State wide average. In 2004 Bowman and Forde each swung by 6% and the swings to the Coalition were also well above 4% in seats such as Blair, Longman, Petrie and Herbert. I suspect Latham went down badly with QLD voters and that the Governments interest rate message in 2004 worked a treat in 2004 in a State where property prices were on the go. Expect a major improvement in Labor’s QLD vote this time.

  5. Sacha says:

    Going the other way, many commentators have said that a similar situation occurs in state elections in South Australia for the Liberal Party. The idea of a systemic bias is interesting – but the number of data points (4) is pretty small, and one has to question the assumption that you can compare elections over 40 odd years.

    “Averaged over this period, Labor needs 50.9% of the vote to win 50% of the seats. Precisely why Labor suffers from this persistent bias is also a topic for another day.”

    Does it make sense to take an average?

    The major point, of course, is that the single member system means there’s no necessary connection between the aggregate fractions of votes won across Australia per party and the share of seats gained. If you want that connection, then you need to count all the votes together (e.g. in a multi-member system).

  6. Sacha says:

    I’ve come to the view that single-member systems are fine (notwithstanding the propensity for local members to be incredibly nimby) as they require a party to have majority support in many different areas of the country to win, and usually a party that gains more than 50% of the 2PP vote wins.

  7. The growing resources available to incumbents in our House of Reps electorates is a factor in this ‘bias’, along with the natural benefit of incumbency. South Australia redraws it boundaries after every election to try to retrospectively remove bias – seems like a futile exercise to me, but maybe their geographical and demographic factors are such that it is felt to be necessary given the inbuilt constraints on democratic fairness contained in single member electorates.

    The only way to ensure a party doesn’t win a majority of seats with a minority of the vote is to have a proper proportional representation voting system. This often coincides with having multi-member electorates, but it doesn’t have to.

    Single member, winner take all electoral systems allow greater bias, although at least we’re not as undemocratic as the USA and the UK which adds the limitations of a first past the post system on top of it.

  8. Sacha says:

    In the USA, there’s usually no pretense of fairness in electoral boundaries – the state legislatures usually directly draw the state and congressional boundaries. The Californian district boundaries, especially around LA, look particularly odd to Australian eyes.

  9. Pedant says:

    Quite apart from what the law says, there are three major problems with the idea that the AEC could be expected to “rectify” so-called “bias” when doing redistributions.

    The first is that it is almost never the case that the whole country is redistributed at the same time, and it’s hard to see what the AEC could do to fix a nationwide issue if, for example, it was only redistributing Tasmania. (And hardly anyone would see much virtue in redistributing the whole country after each election.)

    The second is that it reflects a misunderstanding about how easy it may be to assess the effects of redistributions. When someone is attempting a partisan gerrymander, he/she seeks to waste the opponent’s votes in oversized majorities, without really caring too much about whether the bias so produced is going to be 8% or 12%. It’s a blunt instrument. But to correct a bias of 1%, without overshooting so as to create a 1% bias the other way, requires a razor sharp scalpel, not a meat cleaver. Basically, swing is too variable and unpredictable across electorates to make such fine tuning feasible, now that gross malapportionment is no longer a problem.

    The third is that there are numerous different ways of defining, modelling and measuring bias, and they don’t necessarily produce the same figures. There’s a big literature, to which Prof Jackman, Gary King and Andrew Gelman, R J Johnston, Graham Gudgin and Peter Taylor, and older political scientists such as David Butler, Joan Rydon, etc, etc, have all contributed.

    A number of these points were taken into account when the Joint Standing Committee on Electoral Matters last looked at the redistribution provisions of the Electoral Act.

  10. Andrew Leigh says:

    Andrew B, members’ resources may be going up, but I don’t think it’s yet showing up in the statistics. My rough calculations on the share of contesting incumbents who re-elected each time don’t show systematic trends, either upwards or downwards. Perhaps there’s some countervailing trend (eg. rising IT penetration makes insurgent campaigns easier), or else it might be the case that the elasticity of votes with respect to dollars is actually quite low.

  11. Sacha says:

    AL, in the last Wentworth campaign, Malcolm Turnbull officially spent ~$650,000 and the ALP candidate spend ~$60,000. While there was the Peter King factor (which I don’t think anyone can actually determine), there was a ~2.5% swing to ALP. (NB: there was a pro-labor swing in many safe Liberal Sydney seats).

Comments are closed.