Will The Future See An Even Bigger Senate Crossbench?

It's a long way from the next Senate election to be talking about the future of the Senate, and also a strange time to be doing it with most observers far more fixated on the goings-on downstairs.  But someone is talking about it, so I thought I'd have a detailed look at what they are finding.  The Australia Institute has released a report that attempts to use polling to predict what the Senate might look like in 2019 and 2022.  (There's also a report in AFR, which was originally paywalled but at the moment I can access it OK.)

That another double dissolution "held now" would so flood the Senate with new crossbench Senators as to make the 2016 result look tame is really not worth contesting, and I'm not going to bother with double-dissolution projections off the current numbers.   What is of interest is that the report claims that even if the next two elections were half-Senate elections, on current polling the size of the non-Green portion of the crossbench would increase.  The report's headline projection for 2019 is an extra two non-Green crossbench Senators, and for 2022 after two half-Senate elections, an extra four.  Based on the averageing of two poll results, the report suggests that after two half-Senate elections, there could be a non-Greens crossbench of fourteen Senators, dominated by One Nation with six.



This seems to run slightly contrary to my expectation that the new Senate system is somewhat more friendly to the very small parties in the case of double dissolutions, and somewhat less so in the case of half-Senate polls.  We saw two cases of parties winning seats off about 3% of the vote in 2016, but in a half-Senate election this is extremely unlikely.  But to the extent that the report is right, it is so because One Nation are no longer polling like a micro-party and are now polling levels similar to the Greens.  Parties that can routinely command most of a quota in most states will do well under the new system.  If they can hold their vote at current polling levels (and that's an extremely large if) One Nation could do even better than the Australia Institute suggests.

I don't think you can really use Senate polling to make predictions years out with any accuracy.  This is best considered as a kind of hypothetical exercise about what current levels of voting intention might mean for the Senate down the track if they last.  It's worth analysing this effort mainly because people are likely to keep using projections of this kind to keep fighting their lost wars against Senate reform.  In fact, what the new system would deliver given the votes polled is just what it should deliver - a vaguely proportional outcome where significant voices are well represented.

Issues With Senate Polling

Senate polling generally has a poor reputation, as a result of which it is not often seen.  The biggest problem with it, in my view, is that a polling question that offers voters a choice from, say, five to ten alternatives, does not correctly represent a Senate ballot paper with dozens of party groups on it.  Parties that are named get overestimated, parties that are not named can be underestimated in this process.

We've seen Research Now poll Senate results before - in 2016 the Institute released a Senate poll on which basis it claimed quite that five to nine non-Greens crossbench Senators would be elected.  In the end that proved a modest figure and there were actually eleven.  Here are the differences between Research Now's polling from February-March 2016 and the 2016 Senate result: (I have combined "Liberal" and "National").

All the named parties underperformed the poll (five of the seven by more than their margin of error) while the combined Independent/Other vote was almost four times what the polling had said!  Some of the surge in the Ind/Other vote could be attributed to factors beyond the pollster's campaign, such as the length and tedium of the campaign, or the full impacts of Labor's attempt to portray Senate reform as a conspiracy against small parties.  But overall that the pollster was severely underpolling unlisted parties was obvious and called at the time.

The other reason Senate polling is so difficult is that Senate modelling is so difficult.  It is far more sensitive to errors of a stray point here and there, and the vote shares of small parties are extremely difficult to sample and calibrate.  If Senate polling is a mug's game even days out from an election, what hope years out?

Issues with the interpretation

Let's say we assume the polling figures are extremely accurate - even then, I still have some issues with the way they have been translated to seats based on an average of the two polls.  My main issue concerns the "Independent/Other" category.  For Victoria when dealing with the two-half-Senate-elections scenario and for every state when dealing with the double-dissolution scenario, the Other vote has been treated as a block vote for a single party.  However past evidence shows that this Independent/Other vote will in practice scatter between a very large number of parties and that it may well be that no party among them will have enough votes to win a seat at a half-Senate election.  (At a double dissolution a few would be bound to get up somewhere, but probably not in every state!)

Thus there is no sound basis for throwing a seat to Ind/Other in Queensland or leaving a seat undecided between One Nation and Ind/Other in Victoria.

I get the numbers for half-Senate elections off the votes as given as follows:

NSW Coalition 2 Labor 2 Green 1 PHON 1 (same as report)
Qld Coalition 2 Labor 2 Green 1 PHON 1 (ditto)
Vic Coalition 2 Labor 2 Green 1 PHON 1
WA Coalition 2 Labor 2 Green 1, final seat between PHON and ALP
SA Coalition 1 Labor 2 NXT 1, final two seats between Greens (favoured), PHON and Coalition

Tasmania also deserves comment.  In the absence of meaningful polling, the authors have assumed the vote in Tasmania would be the same as last time.  In fact, with Jacqui Lambie herself not being a candidate, it probably wouldn't be.  In the Section 282 recount (a simulation of an election for six Senators) Lambie finished about 11,000 votes ahead of Lisa Singh.  But it is known that around 5,000 of Lambie's votes did not flow to the rest of her ticket (Lambie had a very high rate of below-the-line personal votes), to say nothing of preferences that would have flowed to her but not to her #2 or, less measurably, votes that would not even have been 1 for her ticket without her in the mix.  So it would be better to treat Tasmania as having a 2-3-1 result in 2019 then a 2-2-1-1 again in 2022.

Thus I get the outcome of the 2019 half-Senate election based on the numbers as given as 15-16 Labor, 13-14 Coalition, 5-6 Green, 3-5 PHON and 1 NXT.  With continuing MPs, 28-29 Labor, 28-29 Coalition, 8-9 Green, 4-6 PHON, 3 NXT, Lambie, Bernardi.  This would not be very different from the current Senate except that Labor would have consolidated at the expense of the minor crossbenchers.  If Labor governs (and they would based on the polling) they would be able to pass legislation supported by the Greens and either NXT or PHON (or perhaps even Lambie).  That's hardly unreasonable as an average of two elections, one of which Labor narrowly lost.

For the combined 2019 and 2022 projections I get the result off this poll as 29-31 Labor, 24-26 Coalition, 10-12 Green, 6-10 One Nation, 2 NXT, Lambie.  Two doses of this medicine in a row leads to a combined majority for Labor and the Greens, and some possibility that Labor would have an alternative path to pass Bills.  The non-Green crossbench is still about the same size it is now, but becomes concentrated in One Nation.

None of this should be surprising.  The poll has One Nation in the same position they are currently in in House of Representatives polling - getting about the same nationally as the Greens.  Perhaps support will collapse by the election or is being overstated even now.  But if One Nation is going to poll 11% on polling day at the next two elections, then they will win and deserve to win several seats even with a normal cycle of half-Senate polls.  That should hardly come as a great surprise to anyone, but in the past this party has struggled greatly to maintain high support for any real length of time.

Lastly, this is the first poll to get any kind of reading on Cory Bernardi's "Australian Conservatives".  For all the fuss made about Bernardi in recent days this particular poll (little known as it is) hasn't picked up on him being any kind of lightning rod for discontented Abbottists.  He's little more than an asterisk in the Institute polling, and doing no better in his home state than anywhere else.  Much more evidence from more reputable sources on this front is needed, but those commentators raving about the supposed Bernardi effect should have commissioned such polling themselves instead of projecting their wishes onto the populace at large.