Background
(if you’re familiar with the electoral systems used in Australia, click here to skip the background)
After every Australian federal election, the Australian federal parliament’s Joint Standing Committee on Electoral Matters (JSCEM) publishes a regular post-election inquiry on the conduct of the election. In the aftermath of the 2019 Australian federal election, the JSCEM has recommended (with dissenting reports from opposition members) the adoption of optional preferential voting (OPV) as opposed to the current compulsory preferential voting (CPV) system.
Australia uses a system of preferential voting where instead of voting directly for a single candidate or party, voters number all candidates in the order in which they would like to see them elected. Under the current system used for the Australian House of Representatives (CPV), voters have to number all boxes sequentially for their ballot to count; leaving out more than one number (e.g. only numbering 1, 2, 3 in an election with 5 candidates), unintentional breaks in sequence (e.g. numbering 1, 2, 3, 5 in an election with 4 candidates) or voting using any other method (e.g. crosses or ticks) will result in the ballot being disqualified from the count. In contrast, under OPV, voters can choose to number as many or as few candidates as they like; as long as the vote has a unique 1 (with crosses and ticks being considered valid 1s in some jurisdictions), it can be admitted to the count.
In Australia, the instant-runoff electoral system, the candidate with the lowest vote total is eliminated at every stage and their votes transferred individually to the voter’s next preference until just two candidates are left in the count. (For a thorough explanation and worked example, I recommend this video) Hence, under OPV, it is possible for ballots to be admitted to the count which express no further preference between the remaining candidates (e.g. a ballot which is only marked ‘1’ for a candidate which is eliminated). Such ballots are referred to as having exhausted, with the percentage of ballots exhausting referred to as the exhaust rate.
Impact of ballot exhaustion
By allowing voters to exhaust their ballots, OPV effectively makes it harder for a trailing candidate to catch up to a leading candidate. For example, let’s say we have a 3-way race between candidates A (40 votes), B (45 votes) and C (15 votes). Under CPV, candidate A would need 73%, or 11 of candidate C’s voters to prefer them to candidate B on their ballot to accomplish a win from a lower first-preference (or primary) vote. (the percentage of voters whose ballots prefer one party over another is also known as a preference flow to that party, as their ballots end up being transferred to that party through the instant-runoff process)
However, under OPV, with candidate C’s voters allowed to not express any preference between A and B, every voter who does so makes it harder for A to overtake B by increasing the percentage of the remaining votes they have to receive to do so (e.g. if 5 votes exhaust, instead of winning 73% they will now have to win 80% [8 of 10] to overtake B). Furthermore, if a large majority of C voters (11 of 15) simply refuse to preference either A or B, then there is no feasible way for A to overtake B and win from behind.
In the current Australian electoral landscape, the centre-left major, Labor, has had a lower first-preference vote than the centre-right grouping (the Liberal-National Coalition, or Coalition for short) but has managed to remain competitive electorally thanks in large part to very strong preference flows (upwards of 80%) from voters for the left-wing Greens. Of the last six Australian federal elections, Labor has only received more first-preference votes than the Coalition just once (2007), yet has managed to basically tie two elections (2010 and 2016) on the two-party-preferred measure and kept things fairly close in another (2019). Under CPV, voters have to number all candidates for their vote to count, while under OPV, voters are allowed to (and often do) refuse to number as many candidates as they want. With Labor heavily reliant on preference flows to save it from a historically weak primary vote, having voters exhaust their ballots is very likely to harm Labor’s electoral prospects.
This has led to a general consensus that a switch to OPV would hurt federal Labor’s electoral chances and boost the Coalition. Generally speaking, OPV is also likely to hurt the chances of minor party and independent candidates, as they might not receive enough preferences to make it into the final two and be elected.
However, I don’t believe I have enough data to model how ballots would exhaust in a matchup between a minor party/independent candidate and a major party.
That being said, considering that Katter’s Australian Party won 2 seats in Queensland under OPV and the Greens have 3 seats in NSW, I doubt that OPV would be as damaging to minor party and independent candidates as commonly assumed.
However, I believe that analyses like this one published in the Conversation overstate the damage optional-preferential voting would do to Labor’s electoral prospects (in particular, even with assumptions unfavourable for Labor, to get the figures for seat changes they did, I had to assume a 60% exhaust rate for all minor party ballots, which from what I can tell has only happened once [NSW 2011] in a landslide election against Labor; this is very unlikely to happen regularly at a federal election). Such analyses tend to assume that a constant share of all minor party voters exhaust their ballots, which doesn’t match up to the data we have from the two states which have had OPV in place recently (NSW and QLD).
In particular, Green voters are less likely to exhaust their ballots than other minor party voters whilst One Nation/Palmer’s UAP/Katter’s Australian voters, I group these three parties together because they tend to have highly correlated preference flows (if one party’s voters move towards preferencing the Coalition, the other parties’ voters tend to do so as well), similar exhaust rates in the election data I have available, and tend to win their highest levels of support in similar (if not the same) electorates.
At elections where more than one of these parties have run, I have grouped them together as “Right-wing Nationalist” (RWN) in my dataset. whose preferences have recently begun to flow more heavily to the Coalition, have historically been more likely to exhaust their ballots than other minor party voters. This somewhat mutes (though it hasn’t yet reversed) the advantage the Coalition is likely to enjoy if OPV is implemented at a federal level.
Furthermore, electorates tend to have very different exhaust rates, which can affect the seat totals enjoyed by either major party. If, for example, exhaust rates tend to be higher in seats which are safe for Labor and the Coalition, and lower (especially Green voters’ exhaust rates) in marginal seats, then it is entirely plausible that OPV could hurt Labor’s 2-party-preferred vote but not their chances of forming government (or even boost the latter). One only need look to the South Australian state legislature for an example of a Labor government holding onto office against popular vote losses thanks to the Coalition’s vote being overly concentrated in safe conservative-held seats.
However, just because electorates have different exhaust rates doesn’t necessarily mean that these differences in exhaust rates remain constant between elections. One could imagine an election where, for example, Labor becomes more environmentalist and therefore exhaust rates in environmentally-conscious seats drop; only for the pattern of exhaust rates to return to normal once Labor rescinds the policy shift. If we don’t know what the differences in exhaust rates will be ahead of time, then there’s no point in attempting to model differences in exhaust rates between electorates; to do so, we need to be able to use some factor to predict the exhaust rates.
Predicting electorate-wide exhaust rates using demographics
(to skip the methods and go straight to what we model would have happened had OPV been in place for recent federal elections, click here)
Here, I use the Census-level demographic data and exhaust data from the NSW and QLD state elections (which were held under OPV) to attempt to predict the differences in exhaust rates of each electorate. Broadly speaking, I first calculate the statewide exhaust rate of each party in a Labor vs Coalition contest (e.g. what percentage of Green ballots in the 2015 NSW state election exhausted before reaching either the Labor or Coalition candidate), then calculate an “expected” exhaust rate based on those figures. (thanks to Rebekah for helping me collate that data)
For example, if the statewide exhaust rate for the Greens is 45%, and the exhaust rate for all other minor parties is 60%, and the Greens won 10% of the vote in an electorate while other minor parties won 15%, then the expected exhaust rate would be 45% * (10%/(10% + 15%)) + 60% * (15%/(10% + 15%)) = 54%. The actual exhaust rate at that election is then subtracted from the expected exhaust rate to get a difference in exhaust rate from expected, which I then run a regression against demographic factors (age, education, religion, occupation etc) to see which factors best explain the differences in exhaust rates amongst electorates. The regression is then used to predict the differences in exhaust rates at a future election, to see if the model is able to predict unseen data instead of merely explaining it (full method below).
- Greens [GRN] (just the Greens)
- Right-wing Nationalist [RWN] (One Nation, Palmer/UAP, Katter’s Australian)
- All other minor parties
Note that there were issues collecting the exhaust rates for the 2011 NSW state election, as the distribution of preferences file listed in the NSWEC website did not have a distribution of preferences for all electorates, but only those where a candidate failed to win an outright majority. Data for the above can be found here.
Next, an expected exhaust rate was calculated for each electorate using the statewide exhaust rates for each grouping and the share of the vote said grouping won in the electorate. We then calculated the difference between the expected exhaust rate and the actual exhaust rate in each electorate.
I then used multilevel regression to model the difference in exhaust rate against the share of the population in each electorate which fell into each of the following Census categories:
- Age bracket (0-19 y/o, 20-24y/o, 25-34y/o, 35-49y/o, 50-64y/o, 65+ y/o)
- Education level (high school, Certificate I-IV, Diploma, Bachelor, Graduate Cert/Diploma, Postgrad)
- Indigenous status
- Language spoken at home (English only, Other languages)
- Religion (Christian, Catholic, Other, No Religion)
- Occupation (categories defined in Census)
I then tossed out any variables which didn’t correlate at at least p < 0.0085 (instead of p < 0.05, I used p < 1 – (0.95^(1/6)) as a cut-off to account for the fact that I’m testing 6 categories) sequentially (i.e. dropping the worst-performing variable until no variables correlated at p >= 0.0085).
The remaining multilevel regression was then run against the electorate demographics for another election, to see how well it could predict exhaust rates.
Obviously, since we don’t have any OPV elections on the horizon we can test against, what I’ve done is develop regressions on old state elections (e.g. QLD 2009 and QLD 2012), then use them to “predict” the exhaust rates of a newer state election (e.g. QLD 2015). It’s not as good as actually predicting an unknown set of exhaust rates (because I might know something about the newer election’s exhaust rates, and subconsciously pick a method which gets me a good result), but in the absence of upcoming OPV elections it’s as good as we can get.
(Exhaust rate data at the QLD and NSW state elections courtesy of Ben Raue)
The factors which best explained differences in exhaust rates seem to be mostly occupational, with a few education variables sprinkled in. So do these variables predict differences in exhaust rates at future elections? These are the predicted exhaust rates (basically adding the predicted differences in exhaust rates back to the expected exhaust rates) for QLD 2015 against the actual electorate exhaust rates:
R2 = 0.49, p < 0.0001
Okay, but maybe that’s just due to the idiosyncrasies of Queensland; maybe the demographic factors which apply there aren’t applicable in the rest of the nation. What if we try to use the QLD model to predict exhaust rates in NSW?
R2 = 0.21, p < 0.0001
The predictive power of a demographic regression is stronger on the higher-quality NSW 2015 data (I was unable to find exhaust rates for electorates where a candidate had 50% of the first-preference vote in NSW 2011, and hence the data is very noisy by comparison):
R2 = 0.41, p < 0.0001
Hence, I think it’s pretty clear that the demographics of an electorate can predict its exhaustion rates in Labor v Coalition contests; and more importantly, that the correlations between a demographic and the differences in exhaust rates remain fairly constant between elections and even across states.
Modelling the impact of OPV on federal elections
To model the impact of adopting OPV on federal elections, I’ve run the demographics regression developed off NSW and QLD’s data on the demographics of federal electoral divisions to estimate the differences in exhaust rates for each electorate. In keeping with findings that Green voters are less likely to exhaust their ballots while One Nation/Palmer/Katter voters are more likely to exhaust their ballots (and all other parties are somewhere in between), I’ve also assumed the following exhaust rates for each grouping (rounded from the average exhaust rates of each grouping at OPV state elections):
- Greens: 40%
- Right-wing Nationalist (One Nation/Palmer/Katter): 56%
- All other minor parties: 50%
This assumption will not hold for every election; exhaust rates can vary wildly from 35% – 60% between elections (though generally the difference between a grouping’s exhaust rate and the statewide exhaust rate remains fairly similar). In particular, experience from the 2015 Queensland state election suggests that if one of the major parties is led by a someone who generates strong antipathy from minor-party voters, exhaust rates can plummet and preference flows shift against them. For example, in the 2010 federal election, where Tony Abbott led the Coalition, would 40% of Green voters decide to exhaust their ballots rather than preference Labor? Or would we see lower exhaust rates similar to those seen at the 2015 Queensland state election where LNP Premier Campbell Newman generated a high level of antipathy from the left?
Would 44% – 52% of Green voters exhaust their ballots federally (as they do in NSW) instead of preferencing Labor, with a less renewable-friendly Coalition at the federal level than at the state level? It’s hard to know without actually having an election, but these numbers are nevertheless the closest we can feasibly get to accurately modelling exhaust rates.
I’ve also assumed that should OPV be adopted, the Coalition parties (Liberal/National) in each state would either merge (a la the LNP in Queensland and the CLP in Northern Territory), or come up with some method of divvying up which seats to contest so as to minimise three-way contests (similar to the Coalition at the state level in NSW). To model this, I’ve assumed that whatever share of each Coalition party’s voters preference the other Coalition party over Labor would simply vote for the Coalition party in their electorate (e.g. in a Labor v Liberal race with a National candidate, whatever share of National voters preferenced the Liberals were simply transferred to them). This probably won’t make much impact (especially considering most three-way races today happen in safe conservative seats) but will probably boost the Coalition’s 2pp by a few fractions of a percentage point.
For my estimates of 2pp, I’ve assumed that the adoption of OPV would cut informal rates by making 1-only, incomplete ballots and ballots marked with ticks/crosses valid. In particular, since NSW has a higher rate of informality compared to the rest of the nation in part thanks to the difference between the OPV system used at the state level and the CPV system used federally, Although from what I can tell, it doesn’t seem like QLD saw a similarly high informal voting rate when it had a OPV system in place. I’ve opted to assume that NSW’s informality rate would be cut to the same share as the next highest among the other states (usually Vic) in my 2pp estimates.
(I am aware of research which suggests that Labor may gain more than other parties if currently-informal votes are admitted to the count, however given that I don’t have exact figures on the differences in incomplete voting between parties, I’ve opted not to include this in my 2pp modelling. It’s not likely to make much difference anyway; such votes tend to be concentrated in safe Labor seats and are likely to have small impacts (a few fractions of a percentage point) on the national 2pp.)
Overall effects of OPV
Broadly speaking, OPV would likely hurt Labor by reducing the flow of Greens voters’ preferences to them. However, the degree to which this happens varies depending on the minor party mix at each election, and the preference flows of the non-Green minor parties. In elections where the Greens make up most of the minor party vote and/or the Labor primary was low, the Labor 2-party-preferred dropped by about 1.5% and Labor lost between 3-5 seats to the Coalition (2010, 2013 and 2016). On the other hand, in elections where the Labor primary was high and/or there were other minor parties in the mix, the 2pp dropoff was more limited, though again about 3 seats would be lost to the Coalition (2007). 2019 would be a mix of the two outcomes; although the Labor primary was very low, a chunk of the 2pp gains made by the Coalition at that election were powered by right-wing minor party voters whose preferences they are less likely to receive under OPV.
Overall, these are fairly small seat differences; in the 28 post-WW2 federal elections, just 6 elections returned governments within a 5-seat margin and off the 6 elections I model here, just one (2010) would have its result changed. Still, it’s interesting to speculate about what might have been, had OPV been in place for said elections. While 2004 and 2013 would remain Coalition thumpings, Tony Abbott’s Coalition would likely gain 4 seats under OPV to form a bare-majority Coalition government in 2010, denying Labor a second term. Would that just bring the Abbott term forward by 3 years? Or would things be different if Abbott had started with the slimmest of parliamentary majorities instead of the comfortable win of 2013 (and a different mix of people in the party room)?
In 2016, thanks to a weak Labor primary and a strong Greens presence with little counterbalancing by right-wing minor parties, the narrow Turnbull win under CPV would have been converted into a comfortable second term in office under OPV, with 51.9% of the 2pp and 5 extra seats. Would Turnbull still have been rolled had he won a decently clear majority? Would Bill Shorten have led Labor into the 2019 election without the impression of having overperformed against Turnbull in 2016?
On a seat-by-seat basis (summary statistics for each election below), OPV would have had some interesting effects on 2pp. For comparison’s sake, I’ve run a simple uniform-exhaust model which assumes every party’s voters in every electorate exhaust at the same rate (48%, average exhaust throughout OPV state elections), which shows how failing to account for differences in exhaust rates can overstate the damage OPV would do to Labor.
Zooming into a seat-by-seat level, the implementation of OPV would shift the contours of the electoral map somewhat. The details for each election are listed below, but broadly speaking the implementation of OPV would likely reinforce what might be termed historical heartlands for both the Coalition and Labor. At every election, excluding non-classic (i.e. Labor/Coalition vs minor/independent) contests, the seats where Labor gains the most on a 2-party-preferred basis under OPV tend to be seats with a long Labor tradition (e.g. Divisions of Hunter, Chifley, Blaxland), where voters may be unwilling to vote 1 for the Coalition but have so far been happy to vote for minor parties and preference the Coalition. Similarly, on the Coalition side of things, the seats they would gain the most in are either seats where they’ve faced strong independent challenges (Divisions of New England and Warringah in 2019) and/or where voters have a historical aversion to voting for Labor but who are happy to support minor party and independent candidates (e.g. Divisions of Mayo, Indi, Wentworth).
On the Labor side, this trend is perfectly encapsulated by the Division of Hunter, which has historically been a safe Labor seat thanks to strong support from mining towns. In 2019, it would have seen the furthest shift to Labor under OPV thanks to the high share of One Nation voters (who tend to exhaust their ballots rather than preference either side), remaining a fairly safe Labor seat instead of turning into a marginal as it did under CPV (and possibly muting its local member’s sudden outspokenness about coal).
(if you’re on a mobile device, scroll right for full data or turn your device landscape. Click the Previous and Next buttons to view all data.)
Electorates with the biggest shifts to Labor under OPV
| Division | Election | 2pp (CPV) | 2pp (OPV) | Difference |
|---|---|---|---|---|
| Hunter | 2019 | -2.98 | -7.56 | -4.58 |
| Kennedy | 2007 | 7.51 | 3.11 | -4.4 |
| Kennedy | 2004 | 8.95 | 5.47 | -3.48 |
| Whitlam | 2019 | -10.91 | -13.51 | -2.6 |
| Kennedy | 2010 | 11.94 | 9.75 | -2.19 |
| Chifley | 2019 | -12.37 | -14.54 | -2.17 |
| Chifley | 2016 | -19.19 | -21.32 | -2.13 |
| Hunter | 2016 | -12.46 | -14.54 | -2.08 |
| Blaxland | 2007 | -18.37 | -20.34 | -1.97 |
| Fowler | 2013 | -16.8 | -18.52 | -1.72 |
| Fowler | 2016 | -17.49 | -19.21 | -1.72 |
| Charlton | 2013 | -9.23 | -10.88 | -1.65 |
| Hunter | 2004 | -13.75 | -15.36 | -1.61 |
| Kennedy | 2019 | 14.51 | 12.93 | -1.58 |
| Hunter | 2007 | -15.92 | -17.48 | -1.56 |
| Chifley | 2004 | -12.98 | -14.52 | -1.54 |
| Hunter | 2013 | -3.67 | -5.18 | -1.51 |
| Macarthur | 2019 | -8.4 | -9.91 | -1.51 |
| Blaxland | 2016 | -19.48 | -20.97 | -1.49 |
| Greenway | 2016 | -6.31 | -7.76 | -1.45 |
On the Coalition side, the twin seats of our former PMs, Warringah and Wentworth, demonstrate how wealthy suburban seats once safe for the Coalition might still remain in the Coalition camp (on a 2pp basis) under OPV.
(if you’re on a mobile device, scroll right for full data or turn your device landscape. Click the Previous and Next buttons to view all data.)
Electorates with the biggest shifts to the Coalition under OPV
| Division | Election | 2pp (CPV) | 2pp (OPV) | Difference |
|---|---|---|---|---|
| Warringah | 2019 | 2.12 | 11.43 | 9.31 |
| New England | 2016 | 16.42 | 25.04 | 8.62 |
| Indi | 2016 | 4.4 | 11.28 | 6.88 |
| Warringah | 2016 | 11.09 | 17.79 | 6.7 |
| Wentworth | 2019 | 9.85 | 16.53 | 6.68 |
| Mayo | 2019 | 2.54 | 8.95 | 6.41 |
| Indi | 2013 | 9.1 | 15.43 | 6.33 |
| Cowper | 2016 | 12.58 | 18.66 | 6.08 |
| Mayo | 2016 | 5.35 | 11.11 | 5.76 |
| Cowper | 2019 | 11.88 | 17.56 | 5.68 |
| New England | 2019 | 17.63 | 23.08 | 5.45 |
| Kooyong | 2019 | 6.68 | 12.08 | 5.4 |
| Indi | 2019 | 12.73 | 17.79 | 5.06 |
| Higgins | 2016 | 10.69 | 15.56 | 4.87 |
| New England | 2013 | 20.71 | 25.34 | 4.63 |
| Mayo | 2004 | 13.59 | 18.08 | 4.49 |
| Mackellar | 2019 | 13.22 | 17.62 | 4.4 |
| North Sydney | 2016 | 13.61 | 18 | 4.39 |
| Barker | 2016 | 15.19 | 19.51 | 4.32 |
| Wentworth | 2016 | 17.75 | 22.03 | 4.28 |
OPV would thus help to keep both sides’ historical heartlands safe from the current realignment of wealthier and more educated voters to the left while less well-off and less educated voters realign to the right, although as the experiences of the UK and the USA show, such change is not easy to resist for long (especially considering we assume that vote shares remain the same under OPV here, while the UK and the US show that vote shares can change in realignments). In addition to shifts among safe seats, the marginal seats that are less likely to flip to the Coalition when taking into account differences in exhaust rates tend to be metropolitan electorates (e.g. Divisions of Griffith, Moreton, Adelaide), which suggests that Labor might be more city-centric under OPV than it is today. Of the marginal seats which would move towards Labor under OPV, just three are non-metropolitan (Flynn, Lingiari, Wakefield):
(if you’re on a mobile device, scroll right for full data or turn your device landscape)
Marginal seats which shifted to Labor under OPV
| Division | Election | 2pp (CPV) | 2pp (OPV) | Difference |
|---|---|---|---|---|
| Blair | 2019 | -1.21 | -2.08 | -0.87 |
| Banks | 2004 | -1.06 | -1.38 | -0.32 |
| Lingiari | 2013 | -0.88 | -1.16 | -0.28 |
| Flynn | 2007 | -0.16 | -0.32 | -0.16 |
| Holt | 2004 | -1.51 | -1.66 | -0.15 |
| Lindsay | 2016 | -1.11 | -1.22 | -0.11 |
| Petrie | 2013 | 0.53 | 0.42 | -0.11 |
| Makin | 2004 | 0.93 | 0.86 | -0.07 |
| Wakefield | 2004 | 0.67 | 0.63 | -0.04 |
I’ve summarised the effects by federal election below, listing the 2pp, electorates which would flip to Coalition under OPV, electorates which would flip if you assumed uniform exhaust but not under our model, and the electorates with the greatest shifts in their 2pp either way per election.
Electorates which would likely flip to Coalition under OPV: Hindmarsh, Parramatta, Swan
Electorates which would not flip under our model: Adelaide
Electorates with the highest swing to Coalition under OPV:
Mayo (4.5%), Bennelong (3.6%), Mackellar (3.5%), Calare (3.3%), Warringah (3%)
Electorates with the highest swing to Labor under OPV:
Kennedy (3.5%), Hunter (1.6%), Chifley (1.5%), Capricornia (0.7%), Scullin (0.5%)
Full data for the 2004 simulations can be downloaded here.
Electorates which would likely flip to Coalition under OPV: Bass, Robertson, Scullin
Electorates which would not flip under our model: Corangamite
Electorates with the highest swing to Coalition under OPV:
Wentworth (3.4%), Mackellar (3.1%), Warringah (2.8%), Curtin (2.3%), O’Connor (2.2%)
Electorates with the highest swing to Labor under OPV:
Kennedy (4.4%), Blaxland (2%), Hunter (1.6%), Chifley (1.4%), Reid (1.2%)
Full data for the 2007 simulations can be downloaded here.
Electorates which would likely flip to Coalition under OPV: Corangamite, La Trobe, Moreton, Robertson
Electorates with the highest swing to Coalition under OPV:
Warringah (3.8%), Mackellar (3.8%), Wentworth (3.6%), Bradfield (3.4%), Lyne (3.2%)
Electorates with the highest swing to Labor under OPV:
Kennedy (2.2%), O’Connor (0.8%), Lalor (0.8%), Hunter (0.7%), Holt (0.5%)
Full data for the 2010 simulations can be downloaded here.
Electorates which would likely flip to Coalition under OPV: Bendigo, McEwen, Parramatta
Electorates which would not flip under our model: Chisholm, Melbourne Ports, Moreton
Electorates with the highest swing to Coalition under OPV:
Indi (6.3%), New England (4.6%), Warringah (4.1%), Mackellar (4%), Wentworth (3.7%)
Electorates with the highest swing to Labor under OPV:
Fowler (1.8%), Charlton (1.7%), Hunter (1.4%), Blaxland (1.1%), Port Adelaide (1%)
Full data for the 2013 simulations can be downloaded here.
Electorates which would likely flip to Coalition under OPV: Griffith, Herbert, Hindmarsh, Longman, Melbourne Ports
Electorates with the highest swing to Coalition under OPV:
New England (8.6%), Indi (6.9%), Warringah (6.7%), Cowper (6.1%), Mayo (5.8%)
Electorates with the highest swing to Labor under OPV:
Chifley (2.1%), Hunter (2.1%), Fowler (1.7%), Blaxland (1.5%), Greenway (1.5%)
Full data for the 2016 simulations can be downloaded here.
Electorates which would likely flip to Coalition under OPV: Corangamite, Lilley, Macquarie
Electorates which would not flip under our model: Eden-Monaro, Griffith, Moreton
Electorates with the highest swing to Coalition under OPV:
Warringah (9.3%), Wentworth (6.7%), Mayo (6.4%), Cowper (5.7%), New England (5.5%)
Electorates with the highest swing to Labor under OPV:
Hunter (4.6%), Whitlam (2.6%), Chifley (2.2%), Kennedy (1.6%), Macarthur (1.5%)
Full data for the 2019 simulations can be downloaded here.
I hope that this post demonstrates that adopting optional-preferential voting at the federal level is not going to “devastate Labor”, as different groups of voters will make broadly different decisions on whether to preference a major party or let their ballots exhaust. In particular, examining the evidence of the impacts of OPV in Queensland and NSW, where Labor has won or remained competitive in most state elections, we can see that failure to model differences in exhaust rates between parties and demographics can overestimate the damage OPV would to Labor’s electoral prospects. As ever, relatively small tweaks in how single-winner elections are run generally have fairly small impacts on the electoral results, and only change the outcomes of elections which are very close to begin with.