How might an election have played out under PR?


On Wednesday I published a blog post arguing that the House of Representatives electoral system needs to change to proportional representation, but I stayed away from any discussion of specific models. Since then, there’s been a lot of discussion of various models, so I thought I would try my hand at a few different models that I think could potentially work.

In this post I’ve run through what I think are the most feasible options to make a change to the electoral system without the need for constitutional change, and what they could potentially look like, both in terms of a map of electoral districts, and the party seat counts based on the voting figures at the 2019 and 2022 elections.

So my preferred electoral system involves multi-member electorates, elected using the single transferable vote. You could also elect multi-member electorates using some kind of party list system and that would probably work fine, although I’m not sure if it would be constitutional.

For the sake of this blog post, let’s assume these electorates use the same voting system as the Senate. I know a lot of PR advocates in Australia are very attached to the specific Hare-Clark voting system used in the ACT and Tasmania, which I also like, but I think the Senate system works just fine.

I’m also assuming that the new system would need to fit within the current Australian constitution, which requires that MPs are “directly chosen”, and allocates set numbers of seats to each state, and prohibits the creation of electorates that cross state borders.

I’m not certain, but I don’t think it’s possible to use an MMP system, as seen in New Zealand or Germany, under our current constitution. I’m also sceptical about whether a party list system with multi-member electorates (of the kind recently adopted for elections to the Senedd of Wales) would be constitutional.

Unfortunately the allocation of specific numbers of seats to each state makes it impossible to set a uniform rule about the number of seats per district. For a start, any PR system would likely result in the Northern Territory, the ACT and Tasmania electing all of their MPs as a single electorate (although I think you could make an argument for maintaining single-member districts where a jurisdiction is entitled to less than three seats, as is the case in the NT).

I think you would set an ideal magnitude, which could be 3, 5 or 7, but in some cases you would need to have a district with a slightly different magnitude to make up the numbers. You could probably write the rules to prioritise lower-magnitude districts in the least densely-populated part of the state, or higher-magnitude districts in urban areas.

There’s no getting around the fact that multi-member districts would cover much larger geographic areas, and would change the nature of local representation. You would still have local MPs, but you wouldn’t just have one, and they’d need to cover a larger area. Proportional representation provides more options to voters in terms of who they vote for: they can vote for someone who lives in their area, but are not as restricted by arbitrary electorate boundaries, and can prioritise representation in other ways.

In order to offset the increased size of districts, I would pair the change in the electoral system with an increase in the size of the House. If the number of senators per state was increased from 12 to 14, that would trigger an increase in the House to about 175 members. That’s a whole blog post on its own, and for most of this blog post, for ease of analysis, I am working on the basis of the existing 151 electorates.

This first table shows the number of seats each state would have been entitled to in 2022 in a House of 175, compared to the actual entitlement, and the number of districts you would need for that number of MPs if your target was districts of magnitude 5.

It's quite convenient that four of the six states have a number of electorates that are divisible by five, but that won't always be the case. In the 151-seat scenario, it's necessary to create one 7-member district and two 7-member districts in Melbourne, in addition to the lower-magnitude districts for the territories.

Indeed if you expand out to 175 seats, the three biggest states neatly fit into multiples of five, but WA and SA now divide by multiples of six. A 6-member district is not ideal, so under my model it would be better to use 5- and 7-member districts where possible. In WA this would mean a 5-member, 6-member and 7-member district.

This next table shows, in the scenario where the House is not expanded, how you would allocate the seats to districts to aim for a magnitude of 3, 5 or 7.

Interestingly, the districts are much more likely to vary in size with Magnitude 7. It's just a number that is harder to fit in. Putting aside the three smallest jurisdictions, which don't have the flexibility to vary in magnitude, there is only two districts with a magnitude other than 3 under the M3 scenario, and just three that vary from a magnitude of 5 under the M5 scenario, whereas there is a need for six lower-magnitude districts under the M7 scenario.

So what could these districts look like? To make this task practical, I have limited myself to combining existing House districts. This does make some of the districts more unusual than would be possible if you were combining LGAs, but this way I can easily combine the results of the election without having to conduct a redistribution, which is what you would do if we were performing this task seriously. It does also mean I can only do it on the basis of the 151-seat House. So bear that in mind: in a real 175-seat House elected proportionately, the districts could make more sense and would generally be a bit smaller.

And of course voting patterns would change under STV. We don't know how this would affect the Coalition, so I have treated them as a single unit, but a more proportional system would likely boost the Liberals relative to the Nationals, since the Nationals tend to hold most of their seats in much safer areas where the Coalition would lose ground, and most non-Coalition seats are contested by the Liberal Party, who would pick up some seats in those areas.

It's particularly difficult to know what to do with independents. Independents can win under STV, but their vote patterns would be quite different. It's also possible it could spur the creation of a teal party of the centre, which probably wouldn't win three seats in northern Sydney but could win seats in more districts.

I've modelled the M3, M5 and M7 boundaries on both the 2019 and 2022 election results, for now just using the House of Representatives results. This is slightly skewed for 2019 since in practice WA should have had one more seat and Victoria one less, but it's still interesting.

This chart shows the results of both elections under the three models, compared to the actual outcome and a pure proportional result for these seat-winning parties (excluding votes for others).

Before I run through the results, it's important to emphasise that an electoral system change is for the long term, and party systems can change, and will change. It would be wrong to assume that these trends are fixed in stone.

STV is not purely proportional, but I don't necessarily think that's a bad thing. The key feature that all of these STV models have in common is that they result in no major party winning sole majority control of the House, with more minor party MPs on both the left and the right, with the exception of the less proportional M3 in 2019, which left the Coalition holding their 77 seats. The Coalition loses less ground to minor parties on its right than Labor loses seats to the Greens.

It does reduce the number of independents, since they tend to have a concentrated vote in one electorate and don't have support in neighbouring seats, but the growth in support in multiple electorates on the north shore of Sydney and in the eastern suburbs of Melbourne mean they do win seats there. Indeed on the M5 model for 2022, I estimate that five of the eight independents elected would have been people who won anyway: Haines, Wilkie, either Ryan or Daniel in Melbourne, and two of the three independents elected on the north shore of Sydney.

Under M3, the Greens win about half as many seats as their proportion of the vote, but it's a high barrier for the smaller right-wing minor parties. I estimate One Nation wins just one seat in 2019 and no seats in 2022.

M5 and M7 are much more proportional. The Greens win about their proportional share under M5 and M7 in 2019. They actual overperform their share of the vote under M5 in 2022. The Greens vote seems to have reached a sweet spot for winning seats in 5-member districts, so when you go to 7-member districts you make it slightly easier, but you also reduce the number of potential districts they could be elected in. This makes sense when you think about the current Senate, where the Greens win a Senate seat in every state under an M6 model, giving them 16.7% of the state senators despite polling much less than 16.7% of the vote. Indeed if you expanded the current Senate to 14 senators per state, the Greens would likely not win a single new seat, although it would make their existing seats safer.

On the right, One Nation do a bit better, winning 5 seats under M5 in 2019, 3 under M7 in 2019, and 2 and 3 respectively under those models in 2022. One Nation would be in the balance of power under M5 in 2019, but that is it. In 2019 under M5, the Coalition could choose between One Nation and a combination of Katter and the three independents. Under M7, One Nation would not have been enough and they would have also required the support of the independents.

What would this have meant for how government is formed? Well, in every PR scenario the 2022 election would have seen a centre-left majority, with the Greens in a strong position to form government with Labor. The Greens would have been strongest under M5, but ultimately the maths is the same: Labor can govern with the Greens or with the Coalition.

In 2019, the Coalition would have held its majority in 2019 under M3, with just one extra One Nation MP on the crossbench compared to the reality. They would have likely formed government either with One Nation or with centrist independents under M5, while under M7 they would have needed both, which would have likely been a messy situation.

Finally, I have made a map! You can toggle through each model and look at the district boundaries, and if you click on a seat you'll see the results for that district and how I allocated the seats. I did go through and allocate all the under-quota seats, many of which were obvious but in some cases were close.

That's about it for now, but I'm sure this won't be the last you hear about proportional representation from me.

Update: Some people have requested a list of which seats were put in divisions so it's in this google doc, along with the results for each division for 2019 and 2022.

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


  1. I was just in the process of replying to someone on your previous article, when this one pop up, so I will just reiterate what I have said there.

    Personally I think too many posters (in the previous article) are tackling this issue from a retrofit perspective. They are all seeking to fit their suggestion into a situation that fits the current constitution (although I haven’t checked what it said), but I find it weird that the Constitution would define set number of seats or proportion of seats per states, which would infer that States don’t grow or decline in population over the years. There is also a lot of discussion around trying to fit seats into a fixed number of representatives.

    I am more inclined to view this from a blow up the entire system and imagine what would be the best way to have political representation in the Australian parliament looking forward for next 100 years. That will most likely mean a Constitutional amendment, but if that is the case and it results in a better form of fairer political representation AND governance – then lets look at it from that perspective.

    I would would also put the future and role of States into that imagining. That is why I was suggesting regions in my previous post. There will be some regions (based on communities of interests) that will have larger population and hence number of representatives. And there will be some regions that will have smaller populations and that is fine, in such instances, they will just have few representatives based on that total population. In the more remote regions, we shouldn’t get hung up with having multiple representative, but maybe just 1 will be suffice (depending on how many houses we land on).

  2. Neil,

    The constitution doesn’t prescribe a specific number of seats but it does prescribe that seats are allocated in proportion to population. So the total number of seats per state would stay the same whether you use multi-member or single-member districts, and thus you’ve gotta fit the magnitude to the seat count.

    You’re welcome to imagine reforms that don’t fit within our current constitution but it’s hard enough to achieve such a change without dealing with our notoriously difficult to amend constitution.


    Sure, I’ll upload a google doc.

  3. I did the numbers for an Additional Member System (similar to Scotland and Wales which would be constitutional because MPs are not elected across State boundaries). In this system 78 MPs would be elected by IRV and the remaining 73 would be top-up MPs elected in a State or Territory-wide electorate.

    With a 5% threshold nationwide or 20% in a jurisdiction the results would be:

    ALP 57; LIB 44; GRN 20; LNP 14; NAT 7; IND 1; CA 1; and KAP 1.

    This would provide a Gallagher Index of 7.44.

    This is pretty high mainly because One Nation was only 0.08% from reaching the 5% threshold and the UAP was 0.90%.

    More detail on this here:

  4. @ Ben. I agree any reform is hard enough without doing the constitutional reform jig, so I will continue on with my imagining for another day.

  5. Qld North 2022 only has 4 winners.

    I’d like to see what the results look like based on senate votes (when you have the full files available).

  6. Actually there are five, but the map only shows the top four parties, and I gave the fifth seat to the Greens ahead of One Nation because of the substantial Labor surplus. One of the more difficult seats to call.

  7. Your districts are pretty good (focusing on M5).

    I’m surprised you put Boothby in SA outer and not Spence (which includes areas like Gawler that aren’t always considered in Adelaide). Same goes for Hasluck in Perth South and Brand in regional WA (Hasluck having semi rural areas in it).

    Of course IRL the districts would be redrawn so it’s a very minor point.

    Victoria’s 8 legislative council districts are a good approximation of what M5 seats would look like (too bad Victoria is 39, not 40)

    As an aside, “Sydney Central” including Parramatta and Fowler threw me for a loop, but it’s accurate. Sydney is not evenly spread. If implemented it would likely have a different name.

  8. Some observations:
    * Independents can still win, no problem (something I was worried about).
    * Every part of the country has at least one coalition and one Labor MP. Thats good for representation, but also means a lot of safe seats. I would hope Robson rotation is a feature of any system.
    * Along similar lines, “Greens member for Hunter” and “Greens member for North QLD” will really help that party diversify its interests. Not sure how the party will handle having a lot of safe seats.
    * NT is probably better served by single member electorates, both marginal, than 2 safe at large seats.
    * A bit concerning that in some areas, huge swings have still led to the same result (e.g. 2-2-1 in Perth South in both elections with 10% swing against Libs).

    You might want to add a “seats changing hands” layer to see where the election battles actually are in this system.

  9. Thanks Ben it’s interesting to ponder this system, would seemingly be a better balance between maintaining some form of local representation while improving the proportionality of the House of Reps. Can’t imagine either of the major parties agreeing to implementing it though – it would likely accelerate the further fragmentation of their vote.

    Under this system I agree that multi-member districts of 5 seats is ideal where possible however there would also be the need to have 3 or 7 member seats as well to be able to match a states quota where it’s not so neat to split in multiples of 5. If possible, a 3 seat minimum for territories could be implemented. I would agree to an additional 2 senators per state, which would increase the Senate to 94 seats, according to my understanding, s.24 of the Constitution requires that the House be as close as practicable, twice the size of the Senate so that would be an expanded House of ~188 seats. So on my calculation (and assuming a minimum of 3 House seats in territories, and 5 in each state) each jurisdiction on September 2021 population figures would return MHR’s as follows:

    NSW: 59
    VIC: 48
    QLD: 38
    WA: 19
    SA: 13
    TAS: 5
    ACT: 3
    NT: 3

    In smaller states like with populations heavily concentrated in the cities the 5 member seats don’t quite work in the smaller states especially where the House remains at ~150 seats. On the current seat allocations the 10 SA and 15 WA seats mean that the 1×5 seat country constituency will also cover a fair amount of outer suburbia. Were the state seat allocations remain the same it would make more sense to have 1×7 seat Adelaide electorate and a 7+5 member Perth – North and Perth – South electorate with a 3 member seat for the rest of each state. In an expanded parliament it would work alot better probably, you could have in SA 2×5 member city electorates in Adelaide and a 3 member rural electorate and in WA 2×7 member city electorates and a 5 member rural/regional one.


Please enter your comment!
Please enter your name here