Senate Reform Performance Review 2019

The results of this year's half-Senate election are all in so it is time to observe how our new Senate system performed at its first half-Senate test.  Australian Senate voting was reformed in the leadup to the 2016 election to abolish Group Ticket voting and the preference-harvesting exploits it had become prone to, and give voters more flexibility in directing their own preferences either above or below the line.  In the leadup to that election, many false predictions about Senate reform were made and were then discredited by the results.  I reviewed how the new system went back then: Part 1, Part 2.  Some of the predictions that were made by opponents of Senate reform concerned the results of half-Senate elections specifically, so now we've had one, it's a good time to check in on those, as well as on how this election compared to 2019.  One unexpected issue with the new system has surfaced, concerning above the line boxes for non-party groups, but it is one that should be easily fixed.



Some issues from the 2016 election don't require any further discussion, including the issue of fixed quotas, the need to get rid of Unweighted Inclusive Gregory for surplus transfers, the need for  external verification that what is in the data files matches the ballot papers and various other suggestions made last time.  I refer readers to part 2 from 2016 for those matters.

Issues to be covered this year include proportionality, what this election might have looked like with Group Ticket Voting, blocking majorities, winning vote shares, the impact of preferencing, just-vote-1 rate, exhaust, informal votes, below the line rates, how-to-vote cards, and non-party group boxes.  I've kept it all in one article this time, but it is a rather long one.

I'm open to requests, though some may be handballed to people with better programming skills or declined if too complex for my abilities.  Before I go any further, enormous thanks especially to David Barry for his brilliant Senate preference explorer (now even more deluxe than in 2016!), Grahame Bowland for his nicely presented verifications of the preference distributions, and Ross Leedham for all kinds of calculations on Twitter.

Proportionality

The 2016 double-dissolution was incredibly proportional, especially after taking preferencing into account in the case of One Nation.  One shouldn't expect the Senate system to deliver a very proportional outcome (especially since we have a preferential system and not a simple primary-vote-proportional system), but it did.  Half-Senate elections are harder, because the division into just six seats per state is just a little bit too coarse for smaller parties.  Even so, the results weren't so bad.  In an election which the Coalition won, it won 19 of the 40 Senate seats on offer, Labor 13, the Greens 6, and One Nation and Jacqui Lambie one apiece.

The following table shows how the parties scrubbed up on seat share compared to national vote share, but I'll always argue that this is a stupid yardstick because of the state-based malapportionment of the Senate and because the territory electorates have too few MPs to generally do anything other than bump up the major parties:

Labor and the Greens have recorded the same number of seats as the Coalition with 1% more of the primary vote, but given the number of right-wing parties getting nothing, it's hard to say they've been short-changed here.  No party with more than 2.5% of the vote nationwide failed to win a seat, and only one party with less than that amount was successful.  That there was such a success is largely down to malapportionment in the case of Tasmania.

The better yardstick for proportionality in the Senate is the average vote and seat shares per ticket per state.  I've trimmed out some of the parties that I only included in the above because they were ahead of Lambie.  Here's the chart on that basis:

I've included the WA and Tas Nationals tickets in the Coalition tally (more dubiously in the case of the former); if they are excluded the Coalition figure is 37.28.

The parties with above 2.5% per state have all won seats, and only one party below that has done so, and only because its support was concentrated in a single state.  One Nation have been ripped off by getting one seat instead of two, but a major reason for that is that Lambie (who has some similarities with One Nation but is on the whole well to their left) has taken the Tasmanian seat that they otherwise would have been very competitive for.

The big three have all overperformed, as is normal because all those votes from micro-parties that don't have enough votes to win a seat have to go somewhere.  If the micro-party voters all preferred all the other micro-parties, then the micro-parties would win many seats, but we know from last time that they don't: preferences scatter, and well-known parties get more than obscure ones.

Among the big three, Labor has overperformed by the least.  At this election in every state the Greens got between 0.6 and 0.9 of a quota, which in all cases put them in the top six on primaries, and their preferencing performance was good enough to stay there.  Labor got between 1.6 and 2.2, which in no case was enough for them to seriously threaten a third seat.  In one case their woeful preferencing performance saw them fail to get near winning a second seat, though they started behind anyway.  It happened therefore that at this election five of Labor's excesses over quota rounded down, but even so their seat tally still exceeded their primary vote share.  The Coalition converted five partial quotas out of six, and only got rounded down in Tasmania.  Labor were rather unlucky, but Labor and the Greens combined were not.

Had this election been a double-dissolution, I estimate the Coalition would have won 31 seats, Labor 25, Greens 10, One Nation 6, and one each for Shooters, Hinch, Lambie and United Australia - naturally a more right-wing Senate than the results of 2016 and 2019 combined.  That would have again been a remarkably proportional result, with all of Coalition, Labor, Greens and One Nation getting a share of state seats that was a few percent above their average state vote share.

What could this election have looked like under Group Ticket Voting? 

Anything might have happened!  The Greens would not have been guaranteed any seats at all for their 10.76%, especially since (as seen in the Victorian upper house) they would have been a poor party to swap preferences with since they would not have had any spare preferences.  If Labor and the Greens swapped preferences, that would have at least ensured a third seat for the left in Tasmania (the only place this would seem to assure it), but there is no guarantee that such a swap would have occurred.  One Nation might have won in every state, or none.  Even assuming One Nation wasn't part of the deal, remaining micro-parties could have dealt with each other to deliver one seat that the Coalition, Greens, Labor and One Nation were locked out of in at least four states; what parties would have won these seats off what possibly tiny vote share through what deals is anyone's guess.  Possibly minor right wing parties (not FACN but maybe those flying a bit more under the radar, eg Australian Conservatives) would have won some seats by snowballing ahead of One Nation.  In some states, it is possible that the outcome would have been two seats each for the majors, two arbitrary micro-party seats and zero Greens.

Blocking majorities

In the leadup to the 2016 Senate election, the wonks within the ALP lost out to the hacks and the hacks decided to oppose Senate reform.  One of the concerns the hacks had was that the system would make it too easy for the Coalition to acquire a blocking majority (38 seats out of 76) against incoming Labor governments by persistently winning slates of three in every state, thereby "forever preventing a progressive Senate", to quote Sam Dastyari.  I pointed out that this was a spurious concern, but the hacks were not deterred.

The concern looked particularly spurious when the Coalition barely retained government in 2016 with a Senate performance that would have given them only two clear slates of three (WA and NSW), with another one possible (Queensland), if repeated at a half-Senate election.  After all if the Coalition doesn't get near a blocking majority when it wins, how is it supposed to get one when it loses?

This election saw a weaker result for parties outside the big three than 2016.  Their average combined vote share per state dropped from 27% to 22.9%, and they won only two seats compared to the five (Hanson, Lambie, Hinch and 2 NXT) they would have won in 2016 had it been a half-Senate election.  Even despite this and even despite the Coalition winning the election with 51.53% two-party preferred in the House of Reps, the Coalition still fell one seat short of getting six three-seat slates in an election that it won.  And getting six three-seat slates at one election wouldn't have been the same as doing it twice in a row anyway.

We should also consider the Coalition's position in terms of Labor-Green blocking majorities.  Had the 2016 election been a half-Senate election, Labor would have carried 12 seats into this one, and would now have 25, while the Greens would have carried 4-5 seats in and would now have 10-11.  It's perhaps one seat better for them than what happened.  Various alternative history rabbit holes concerning Lambie not running in both 2016 and 2019 are possible, and maybe there's one of them where Labor gets an extra seat without any negatives, but in no case do Labor and the Greens get a blocking majority out of one scraped and one modest Coalition win.  I still do feel that more minor right parties need to start merging to make better use of the system for their side of politics and combat the success of the Greens at this election, but if they won't, then that's their problem.

Winning Vote Shares

Of the 36 state seats, 22 were won on raw quotas, in one case (Tas) complicated by some of the votes being held down the ticket and delivered late in the count.  13 seats were won on partial quotas of between .611 and .934 Q; every candidate with a notional partial quota over .6 Q won.  The only overlap between winning and losing was that the Coalition won with a spare .513 quota primary in Victoria (where they were in sixth place and no-one could catch them) and Labor lost with .580 quota primary spare in Queensland (where they were seventh and went backwards on preferences.)  In theory it's possible to win off 0.3-0.5 Q in half-Senate elections in this system but there were only three totals in that range (One Nation in WA, NSW and SA) and the breakdown of other parties meant they weren't competitive.

Preferencing Impact

None of the final seats this election ended up being even remotely close, and in every case the top six candidates on primary quotas won.  However, there were some changes in the order and some useful pointers to possible future results.  In WA the Liberals started 1% behind Labor and beat Labor for fourth by 0.1%.  In Queensland One Nation started .06% behind the LNP and beat the LNP by 1.97% (again showing that One Nation wins seats despite exhaust, not because of it.)  For sure, One Nation would have outpolled the LNP on primaries but for competition from One Nation defector Fraser Anning.

There was some speculation that Derryn Hinch might put in a gallant charge from a small primary in Victoria but the gap soon became clearly too wide and in any case he went backwards compared to the Coalition.  He did however overtake One Nation.

Just-Vote-1

If a voter numbers just one box above the line their vote, while contrary to the instructions, is saved by the savings provisions and counts for the party they have chosen only.  At the 2016 election there was a vaguely prominent just-vote-1 campaign from Sydney shockjock Ray Hadley.  This election there was none of this so how did we go?  (Percentages are ATL #1 out of all votes whether above or below the line.)

The overall rate of just-1s went up everywhere except Queensland.  In NSW, this was to be expected as a result of increased contagion from the very recent state election, which has fully optional preferencing in both houses.  In Queensland, the drop may have been caused by the switch from optional to compulsory preferencing at state level - nonetheless the low rate in Queensland is highly impressive.  Overall I'm not sure to what extent the modest rise should be put down to the lack of a large education campaign like that seen in 2016, and to what extent to gradually increased awareness that ATL just-vote-1s would be counted.  (Note: these figures also include votes that exhausted after 1 for other reasons.)

The lowest just-vote-1 division in the country was Franklin (0.89%) ahead of Clark (0.92%) with Ryan third on 0.98%.  But Franklin's above-the-line rate is only 69% and Clark's 63% whereas Ryan's is 91%, so Ryan has the lowest by proportion of all ATLs.  (In 2016 Clark, then Denison, won on 0.85% from Franklin 0.86% and the rest were over 1% with Fenner third.)

Exhaust

It was fairly common to see predictions that exhaust would be much higher in the half-Senate elections than in the 2016 full-Senate elections.  In fact, this wasn't the case, though results varied by state.  The exhaust rates from the 2016 elections were complicated by needless throwing of preferences past the point where outcomes were beyond doubt in some states' distributions, so I calculated an effective exhaust rate of 5.1% for 2016.  This issue didn't resurface in 2019, and the exhaust rate this year at the point where the result was beyond doubt dropped slightly to 4.8%.  Exhaust rose in Victoria (5.2% to 7%) and South Australia (2.0% to 2.3%) and fell in NSW (7.3% to 5.6%), Queensland (4.2% to 3.9%), Tasmania (2.8% to 1.9%) and WA (3.6% to 2.0%).  It was 0.1% in ACT and zero in the NT.

[Note added 17 Sep 2019: I have corrected WA and SA figures to align with The Tally Room as the final exclusions in WA and SA were irrelevant to the outcome.]

Informal Votes

Informal votes rose half a point to 5.54% in the House of Reps, which is a matter of some concern, especially as it appears the Senate system could be contributing to the rate in divisions with 8 or more candidates.  However, in the Senate, informals dropped by 0.13% to 3.81% (see past trends in the 2016 article).  Changes by state/territory were minimal.  The informal rate looked higher earlier in the count because it was common for votes requiring the use of savings provisions to be counted as informal in initial counting and corrected later.

Below The Lines

Again see last time for some history of this issue.  In comparison to 2016, factors that could have driven the BTL rate upwards included the smaller number of candidates.  However this was offset by a smaller number of incumbents, and in some states there were very few competitive incumbents.  A number of the best known names were elected for full terms in 2016, and a couple of very prominent ones (Scott Ludlam in WA and Nick Xenophon in SA) have since flown the coop.

The rise in NSW is down to Jim Molan, who received 42.6% of his state's BTL vote (possibly a record).  I have written a full article about Molan's result.  The fall in Tasmania was mainly down to there being only one Liberal incumbent.  The ACT's rate is volatile and has been higher in the past when there were very few candidates on the ballot, but the presence of Anthony Pesec as a significant candidate with a blank ATL box contributed to it (see below).

BTL candidate effects did not affect the winners list in this election.  Lisa Singh's vote in Tasmania (5.8%) declined slightly from her winning total in 2016 but she was unable to win because Labor's ticket vote was too low and Jacqui Lambie's vote increased.  Singh eventually finished eighth, but had she got over One Nation, Lambie would have cleared quota easily on One Nation's preferences anyway.  Given Labor's poor performance and Lambie's strong result, Singh would have needed over 10% to win.  Singh's presence did have the effect of delaying Catryna Bilyk's election for long enough for Nick McKim to cross the line ahead of Bilyk, but order of election has no significance in a half-Senate election.

In NSW Molan's vote simply resulted in him remaining in the count until being excluded in tenth place, with most of his vote flowing back to his ticket, although 20% leaked to One Nation.  In all, nearly 28% of Molan's #1s had #2s outside the Coalition ticket, led by McCulloch (ON, 8.4%) and York (Conservatives, 6.4%).  Molan's result does suggest some optimism about a BTL candidate sometime winning in a mainland state at least at a double dissolution, but expect a crackdown on such behaviour within split Coalition tickets in view of its threat to Coalition seat balance.

How To Vote Cards

In 2016 we saw that Senate how to vote cards were little followed, with Coalition cards getting about 30% follow rates, Labor about 14% and the Greens about 10% (all lower in Tasmania).  Right-wing minor parties sometimes approached 10% while left-wing minor parties often managed only a few percent or less than 1%, even in one case actually zero.

Not much has changed.  This time, the Coalition didn't break 30 anywhere and managed only 8% in Tasmania and 10.2% in NT, with other rates from 19.7% (NSW) to 28.4% (WA), the highest for any ticket anywhere.  Labor managed only 4% in Tasmania (because of the Garland issue, see below) and otherwise ranged from 15.8 to 18.8.  The Greens ranged from 7.1 (Queensland) to 14.4 (ACT) with a fairly high 13.3% in Victoria.

One Nation ranged from 0.7% to 2.4%.  United Australia's card was impossible to follow in the ACT and followed by only 0.3% in NT and 1.2% in Tas, otherwise ranging from 3.4% (NSW) to 5.8% (Vic).

Separate Nats cards were followed well in Tasmania (15.5) and WA (15.9) but apart from that there were only two more cases of a micro-party breaking double figures: Australian Christians (WA, 15.6) and Tony Moore (Qld, 10.8 - but with his ATL vote doubtless deflated for the reason noted below).

Beyond this, micro-parties tended to get follow rates from below 1% to sometimes 6-7%, without that much pattern by wing of politics.  Crossing between different branches of right-wing politics (CDP to Shooters, One Nation to Liberal Democrats) tended to cause especially fast drop-off rates.  Special mention should be given to:

* The Pirate Party's NSW card, which ran from 1-10, but of the 16 voters (0.1%) who made it to 6, 15 jumped ship immediately (1 to ALP, rest to exhaust) with only one going the distance.  The Pirates' cards did not exceed 0.3% follow rate in any state.

* The Small Business Party, which produced bizarre party orderings followed by 0.1% (NSW) and 0.08% (Vic), the latter being the lowest follow rate I could find.

* The Citizens Electoral Council, whose card was followed 1-6 by one voter each in NT and Tasmania.  (The NT card preferencing Rise Up Australia 2 and Greens 3 was obviously crackpot, but nine voters still made it that far, only for seven to run for the hills when preference 4 went to a blank group box.)

* Yellow Vest Australia, who produced long-form HTVs with every ATL square numbered as well as a short-form option for those with less time (who were insulted for not taking voting seriously).  35 voters followed the long-form ones all the way in Victoria and WA combined, only slightly fewer than went for the short-form versions.

A Jim Molan (Lib) BTL how-to-vote card 1-12 was followed by 17.5% of Molan's voters.

Blank ATL boxes for non-party candidates

This election exposed one unintended issue that deserves fixing.  Candidates who don't want to, or don't have time to, register a party can run as grouped candidates with an above-the-line box.  An above the line box is critically important in the new system as one cannot receive any above the line votes or preferences based on ATL votes without it.  When a non-party ticket runs with such a box, the box is not labelled.

In 2016, nobody significant ran with one of these blank non-party ticket ATL boxes, but in 2019 three significant candidates did - Anthony Pesec (ACT), Craig Garland (Tas) and Hetty Johnston (Qld).  Garland and Johnston recorded poor primary vote results relative to their public profiles, while Pesec at least managed to get his deposit back.  However there were many anecdotal reports of the blank boxes for Pesec especially, and also Garland, causing voter confusion.  Some ACT voters reported being incorrectly told by AEC staff that they could only vote for Pesec below the line.

All three candidates recorded a very high share of their vote in BTLs - 48% for Johnston, 66% for Pesec and 80% for Garland.  This is something that is not new - if one looks at the 2010 results for prominent candidates Stephen Mayne and Cheryl Kernot one sees the same thing.  However, those candidates could still get full preference flows from Group Ticket Voting, even if their low primary votes made it far more likely they would be excluded early in the count.

This election has showed starkly that blank ATL boxes confuse voters to such an extent that many voters don't preference them even if they might approve of the candidates.  A large amount of evidence was published on the button press thread (see ACT and Tas), to which I can add some more depressing examples:

* In the ACT, above-the-line Greens voters "preferred" the Liberals to Pesec 62.4-35.9, and above-the-line Labor voters "preferred" the Liberals to Pesec 64.6-32.3. (This was typically because they left Pesec's box blank.)  But below the line voters for the Greens' Penny Kyburz preferred Pesec to the Liberals' Zed Seselja 85.5-13.3 and those for Labor's Katy Gallagher did so 75.1 to 22.7.

* In Tasmania, Garland was preferenced #2 above the line on both Labor and Green how-to-vote cards.  However only 12.3% of Labor ATL voters preferenced Garland anywhere at all, including last, and only 24% of Greens voters - who one would expect to be able to work this stuff out especially in Tasmania - did.

It's obvious that blank group boxes greatly confuse voters because they don't "look right".  This isn't necessarily fatal, since Nick Xenophon got elected with one in 2007 (though I am unsure if the rules have changed since then).  However, a candidate would have to do a vast amount of publicity to prevent the confusion.  This should be reformed by allowing some kind of generic label that makes it clear the group is not a party.  A suggestion would be, for instance, "Group O (Garland/Duncan)" (using the first two surnames in the group even if it has more than two.)  Anything that makes it look like a regular box - oh and not "independent", because they may not be; they may be just members of a disorganised party.

Other items may be added if there's anything else of interest.

------------------------------------------------------------------------------
Donations welcome!

If you find my coverage useful please consider donating to support the large amount of time I spend working on this site.  Donations can be made by the Paypal button in the sidebar or email me via the address in my profile for my account details.  Please only donate if you are sure you can afford to do so.

------------------------------------------------------------------------------