You can’t just legislate ‘fairness’ in our voting system


The South Australian Liberal government failed on Tuesday in its attempt to restore the “fairness clause” which used to apply to SA state redistributions. The clause was an attempt to ensure the “correct” winner in state elections by requiring the redistribution commissioners to draw boundaries that would give a majority of seats to a party that won a majority of the two-party-preferred vote.

The clause was introduced by a state Labor government via a statewide referendum in 1991, and was then abolished by the outgoing state Labor government in 2017, shortly before they lost the 2018 state election.

The clause is based on a false premise: that you can require “fairness” between major parties in a system of arbitrary winner-take-all contests which is by its very nature unfair. If you want a fair voting system, you need to look to proportional representation systems.

Since 2002 there have been three elections where Labor has won power despite losing the statewide two-party-preferred vote.

Mike Rann’s Labor government won power in 2002 with 23 out of 47 seats, and just over 49% of the two-party-preferred vote. Labor then won the 2006 election comfortably, but then won a solid majority in 2010 despite losing the two-party-preferred vote. In 2014, Labor fell one seat short of a majority and managed to stay in government, despite polling only 47% of the two-party-preferred vote.

The Electoral Districts Boundaries Commission has tried valiantly to enforce the fairness clause (with the exception of the redistribution before the 2014 election, where the EDBC effectively gave up on trying to meet the fairness criteria), which has led to quite dramatic boundary redrawings, but the clause is based on a false premise.

Not every seat will swing in the same way at an election. So you can’t predict how a statewide vote total would play out based on the previous election’s pendulum. Not only that, but differential swings are often systemically consistent: one party will do better in marginal seats potentially due to a swathe of MPs with new personal votes or a better campaign mechanism. The lack of uniformity does not wash out.

All electoral analysts understand this when we look at a classic pendulum. It’s a useful tool but it is a blunt instrument. It certainly is not precise enough to be legislatively mandated to dictate electoral boundaries.

The fairness clause led to much larger proportions of the state’s voters moving between electorates at every redistribution, breaking the links between MPs and their constituents. It often punished successful MPs who had built up local popularity with a dismembered electorate. It also required the commission to prioritise political balance over numerical balance, leading to smaller numbers of voters in some electorates than others.

It’s impossible to produce a “fair” result under a single-member district system, because it is fundamentally not a fair system. It consistently under-represents minor parties, and the seat balance between major parties is only tangentially connected to their relative support. One party can consistently overperform their vote if their voters are distributed in a more efficient pattern around the state. If one party has a handul of electorates with massive majorities while the other has smaller majorities spread over more seats, the latter party will usually win close elections.

We see this in the case of the US House of Representatives. While there is a shameful gerrymander which has boosted the Republicans, they have also benefited from the sorting of voters which has given Democrats some colossal majorities in urban centres. In the case of South Australia, the Labor vote is simply more evenly distributed.

The fairness clause also relies on the assumption that the voting system is dominated by two parties. Imagine that Nick Xenophon’s SA Best had won a sizeable number of seats, which briefly appeared possible. How would you ensure a “fair” single-party winner when three parties are winning significant numbers of seats? You simply couldn’t do it.

You can’t have a fair result in the single-member-district system. This system by its nature under-represents minor parties and tends to produce single-party majorities. If you want a fair result, you should instead look to a proportional representation system. That would do a better job of representing the Liberal and Labor parties fairly, but it would also bring in other players. The current Liberal government wants to have its cake and eat it too: they want the system to require relative fairness in relation to Labor while locking out other players. It doesn’t work like that.

Liked it? Take a second to support the Tally Room on Patreon!


  1. I agree that a “fairness” clause like this is ridiculous, but I’d be interested to know whether TPP preferred swings have become increasingly less uniform since say the early 90s, when this clause was introduced.

    Anecdotally, it seems on the Commonwealth level swings are becoming less uniform, but I haven’t crunched the numbers so I may be talking garbage.

  2. Does anyone have details on how Labor initially got around the entrenchment of the fairness provision? I realise that putting legislated measures beyond the reach of a later parliament is very difficult to do, but I actually thought the mechanism that had been used in SA (essentially creating a cause of action if the provisions – including the entrenching provisions – were changed without a referendum) was quite clever and couldn’t figure out how they were going to bypass it. But it was pretty quickly conceded that it could be done once they had the numbers.

  3. Ignore my previous comment. I figured it out.

    The entrenchment provision (section 88 of the Constitution Act) does not require a referendum unless the amendment touches certain issues, which do not include the (former) fairness provision.

    Amazing what coming back to an issue with fresh eyes can do for comprehension.

  4. Clinton

    The frequency distributions of TPP swings and TPP primary votes at a Federal level are not normal distributions (leptocurtic and skewed) when viewed en masse 1961-2019, but the standard deviations of these parameters are too large to see trends in the statistics between two separate elections.

    Election Abs swing Skewness Kurtosis SD of Swing
    1961 4.31% 1.43 3.10 3.41%
    1963 4.31% 1.43 3.10 3.41%
    1966 5.81% 1.47 2.74 4.71%
    1969 7.80% 0.52 0.29 3.75%
    1972 4.22% 0.49 -0.36 2.41%
    1974 2.92% 0.97 0.31 2.20%
    1975 7.25% 0.44 1.77 2.65%
    1977 1.96% 1.23 1.37 1.64%
    1980 4.47% 0.74 0.75 2.79%
    1983 4.04% 0.51 -0.28 2.25%
    1984 2.05% 1.24 1.77 1.50%
    1987 2.11% 0.99 0.92 1.60%
    1990 3.17% 2.11 8.03 2.81%
    1993 2.87% 1.43 2.37 2.49%
    1996 5.13% 0.54 0.07 3.34%
    1998 5.41% 1.96 7.24 3.63%
    2001 2.82% 1.06 1.21 2.12%
    2004 2.98% 1.29 2.71 2.24%
    2007 5.66% 0.47 0.49 2.85%
    2010 4.76% 2.44 8.74 4.27%
    2013 4.78% 2.45 8.81 4.26%
    2016 10.60% 0.82 -0.17 8.73%
    Means 4.52% 1.18 2.50 3.14%

  5. If I were the SA Liberals, I would be considering either an additional state-wide district (either single member (in combination with an evened number of local electorates) or multi-member (still majoritarian, like the pre-1949 windscreen wiper Senate)) or 2PP MMP (with proportional allocation based on the 2PP vote, not the primary vote).

    A slightly modified 2CP based MMP, where any party that gets into the 2CP in one or more seats (not unlike the winning a certain number of seats bypasses the threshold in most MMP systems) gets included in the proportional section, might be more likely to get parliamentary support. Such a system would likely often lead to 3CP MMP, with the Greens being the third party.

  6. Clinton I don’t think you are talking garbage. Voters have become increasingly unwilling to follow Party HTV leading to strange results.
    Looking at both Federal and Queensland marginal seat results post election and prior to declaration I have been wrong on many occasions with my predicted results.

    I find it totally inane for a voter who voted for a minor Conservative party to place Greens above ALP (or vice-versa) but it happens even when HTV suggests otherwise.

    Preference votes go all over the place and I suspect that we’ll over half are donkey votes once first few preferences allocated. This is a mixture of laziness and a recognition that once a preference has been allocated to a major party that it matters not one iota how rest of preferences are allocated. Of course exceptions to this can be found singly where minor party candidates are possible winners ( Kennedy Indi or Melbourne) Voter non compliance with HTV suggested order of course is impacted by whether HTV get into voter hands. Many minor parties have insufficient volunteers to man booths. There is no doubt that the mentality of a person who votes for minor conservative, independent or Green results in an unwillingness to do as suggested/ directed by a HTV.
    There is no doubt that voters seperate Senate and House votes and a good percentage vote differently in each house.

  7. I agree that trying to enforce fairness across single member constituencies is an extremely difficult task. And it’s true that to insist on the final seat share reflecting the overall vote share is really an argument for proportional representation.

    But I also think that even within the constraints of the single member system, it is reasonable to attempt fairness. Given that it’s trivial to draw a gerrymander that favours one side or the other, it ought to be possible to draw an anti-gerrymander that favours neither side.

    Community of interest should always be the primary criteria when drawing districts; otherwise what’s the point of having separate districts? But the overall electorate never divides evenly into equal sized districts. There are always judgement calls.

    The court drawn map for Pennsylvania’s congressional districts is instructive. The court was forced to redraw the boundaries because of the unconstitutional Republican gerrymander. But it appears that instead of drawing a strictly apolitical map, the court instead went for a politically balanced map. The natural lay of the land tends to favour Republicans, so the subjective decisions – e.g. where county boundaries had to be crossed – tended to favour Democrats. This minimised the natural Republican advantage. The following article, which goes through the redrawn districts one by one, is instructive:

    The same is true of South Australia. That the Liberals won in 2018 with a 52-48 result, after losing in 2014 with a 53-47 result, was in large part due to the ECSA’s decision to take the fairness criteria seriously.

    We should also be clear about the conditions necessary to satisfy fairness. The most obvious precondition is a strong two party system. That obviously holds true in the US. It’s reasonably true in Australia, but not completely. It should not be forgotten that SA Labor’s minority vote in 2002 and 2014 only gave them minority government. Had the Liberals not lost typically safe seats to independents, they would have won those elections.

    The other thing is there needs to be a clear, consistent test of the political leaning of each seat. Single member results don’t work so well because sitting member factors are real. That’s why American analysts use presidential results rather than congressional results. Perhaps ECSA should have been using LegCo figures instead of LegAss figures?

    There’s also arguments over what “fairness” means. In the US, these debates get bogged down with arguments over proportionality, or “competitiveness”, or minority representation. To me it’s simple: 50% of the two-party vote should result in 50% of the seats. (a.k.a. the mean/median test.) In addition, the most marginal seats should be evenly distributed either side of that 50% mark. (i.e. on a bell curve.) That’s it. That’s all fairness can mean in a single member two-party system.

  8. As David said above, when the SA electoral Commission finally got serious about applying the fairness law rather than just tinkering at the edges as they had previously, the result was immediate – the government changed. This is strong evidence as to the falsity of claims that boundaries can’t be made fairer. After all if boundaries can’t be made fairer (or less fair) it would mean that the concept of a gerrymander is invalid.

  9. Interesting long term analysis of Decline in ALP vote The Australian 7 Nov 2019 p 4.
    State by state declines 2007-2019

    NSW -9.5%
    Vic -7.83%
    Qld -16.23%
    WA -7%
    SA – 7.8%
    Tas -9.36%
    ACT -10.0%
    NT -5.38%
    Anthony Chisholm cites Decline in Bowen as being symptomatic of problem faced by ALP
    2007 57%
    2019 20%

    Bowen used to be the reddest town in Australia electing Australia’s only Communist MP.
    In NSW and Victoria a lot of loss in Primary vote returns to ALP as preferences but not the case elsewhere.
    Palasczuk had better do some hard work in next few months outside the Brothel Zone of South Brisbane. In fact sensible thing would be to move Jackie Trad out of South Brisbane concede it to Greens move he to Townsville and concentrate on winning 95% of population who think jobs and prices are more important than climate Rebellion and protection of Sodom.

  10. ALP Review of Election supports my previous comment. ALP Review is an excellent document examining good points and bad points of ALP campaign. Whilst I disagree with a huge amount of ALP policy the Party is the most professionally run party in Australia. Makes Coalition and minor parties look incompetent at times.
    Unfortunately Party not judged on professionalism only on results.

Comments are closed.