Monday, February 29, 2016

Senate Reform: JSCEM Submission Take Two

I'm posting links here to my submissions to the warp-speed JSCEM inquiry into the Commonwealth Electoral Amendments Bill (see Senate Reform: It's Finally On! for more comments).  They are also available as a single document on the JSCEM page.

Main submission
Simulation - Half-Senate Elections
Simulation - Double Dissolutions (See correction further down to Vic 2007 double-dissolution simulation)

After spending a lot of time looking at the Bill over the past week my view remains that it is a massive improvement on the broken Senate voting system we currently have.  However it appears to have significant flaws in the treatment of below-the-line votes and these should be either justified or remedied.

What has become apparent in the brief time available for debate about the current Bill is that by encouraging voters to fill more squares above the line (a minimum of six instead of a minimum of one) the Bill makes deciding what to do with below-the-line votes more difficult than in the original JSCEM model.  The original JSCEM model allowed a voter to just vote 1 above the line, or to direct preferences above the line, and it was very safe to conclude (based on the Victorian model) that the rate of below-the-line voting in that system (with a minimum of six boxes) would have been low.

However, because the new model will ask voters to fill in six squares above the line, if the voter is only asked to fill in six squares below the line, then below-the-line voting becomes more attractive.  A lot of us would like to see this in the long term, and see voters have more effective choices within parties as well as between parties, especially given the woeful Senate preselection decisions made by major parties from time to time.  But springing into that world right away without thinking about it very carefully could produce some problems, which I outline in my submission.  Basically the Senate doesn't have a tradition of intra-party contests or personal-profile politics, and to spring too much of that on Senators at short notice could lead to some funny results.  It should happen eventually, but not  in my view just yet.

In the meantime I'm thinking - tentatively - that the best model to deal with below-the-line votes is to ask voters to number from 1 to 12.  Antony Green has independently formed the same view.  The reasons are outlined in my submission.

My submission includes a complete resimulation of all Senate elections from 1990 onwards, firstly on the assumption that the same votes were cast at half-Senate elections, and secondly assuming that the same votes were cast at double dissolutions. System change causes change in who contests an election and what sort of vote they get - as does calling a double-dissolution even when there is no system change - so to some degree this is an academic exercise.  However it does show that the Coalition would not have been near winning a majority under the proposed system except at the election (2004) at which it actually did.  If anyone spots any errors in the simulations let me know.  They were all done in one day and I have not had time to thoroughly recheck them - but any mistakes will not change the overall picture.

The simulations show that a double dissolution in 2013 under the new system would have been a crossbencher picnic, electing about 12 non-Green crossbenchers.  Under the current system the number would have been even greater.  While PUP, which would have won about five seats, has since imploded, the simulation does show that crossbench wins are relatively easy at double dissolutions under the new system.  It's probable a DD "held right now" would reduce the number of parties on the non-Green crossbench, but it might not reduce the number of crossbenchers greatly.  We also cannot assume that just because Ricky Muir only got 0.5% of the vote last time, that he would do so again.

As with my analysis of the previous JSCEM model, I find that the combined results of the 2010 and 2013 half-Senate elections under the proposed new system would have given Labor and the Greens a blocking majority in the current Senate.  It may seem from this that the proposal is too kind to these parties, given that the 2010 election was a draw and the 2013 one was a thrashing for the ALP.  However, a major part of the reason why an ALP/Green blocking majority would have been possible is the terrible 2010 result for the Coalition in Tasmania.  Tasmania's large share of Senate seats for its population distorts the proportional nature of the Senate, so that trying to read an expected Senate result from a national 2PP doesn't work well.

I am attending (via phone hookup) a round table at JSCEM tomorrow and perhaps I will have more to say after that, especially on the practical aspects.

Update: Below The Line Error Rates

The Guardian reports the Coalition does not intend to allow further below-the-line changes, although there is still no coherent justification given for not doing so.

Special Minister of State Matthias Cormann is however said to have stated that "by 2013 federal election figures the proposed improvements to the savings provisions “would have saved about 45% of informal below the line votes”."

I think this is extremely unlikely to be accurate as quoted, and wonder if the quote is truncated.  In the 2001 formality survey of below-the-line votes (the most recent available to me) it was found that excessive repetition or omission accounted for only 2.87% of all informal Senate ballot papers, or 6% of all "below-the-line errors".  Some of these would be saved by the new provision and some not.  Despite the increase in the number of candidates since 2001 (candidate numbers have nearly doubled) I find it very unlikely that omissions and repetitions have swollen from 6% to 45% of below-the-line informals, let alone that informals saveable by the change would have risen to that much.  Senate informal voting generally has gone down, but the leading causes of BTL informality probably wouldn't have.

It may well be the case that the change would save 45% of those votes that are currently informal as a result of excessive omissions and repetitions.  In 2001 that would have been 45% of 2.87% of 3.89% of the total vote, ie one vote in every 2000 - a gain that could well be lost if voters became more careless because they had heard they had more room for error.

[Update: based on Tony Nutt's comments to JSCEM it appears my suspicion was correct.  Nutt stated that informal BTLs were only about 2% of BTLs (or 8500 votes), which the new savings provision sought to drop to below 1%.  The suggestion is that the rate of votes saved would be even less than my 1 in 2000 estimate above, possibly reflecting higher accuracy in BTL voting as a result of online preparation tools.  But at an actual election under the new system it will be even lower, because many voters who vote BTL now will take up the ATL preferencing option.]

Update: Labor's Modelling

I've seen a scarcely believable report in the West Australian about how Bill Shorten got David Feeney to resolve the impass between Gary Gray and Stephen Conroy over the original JSCEM model (not the current one).  It contains some very silly stuff (not all of it Feeney's):

"The application of above the line OPV to a hypothetical scenario where, over two election cycles, a result modelled on the 2013 election is repeated twice".

Well hello, if you're so hopeless as to lose two elections 53.5:46.5 in a row you should hardly be surprised that your opposition gains control of the Senate.

"Under this scenario, the Senate ballot is only contested by the ALP, Coalition and Greens, all voters cast a simple "1" above the line, and the 2013 vote shares of all other parties are re-allocated to the ALP, the Coalition and the Greens according to observed preference flows at the 2013 election"

Firstly you're assuming we have OPV but no Labor or Green voter uses it (despite all evidence to the contrary from HTV-card preference flows in the NSW and Queensland state elections.)  But more importantly, you're assuming that every non-Green crossbencher loses at two consecutive Senate elections.  The falseness of that assumption is exactly why it's so difficult for the Coalition to get control of the Senate.  We have to go back to 1955-8 to find the last time only one crossbench force won seats at a Senate election twice in a row (it's very rare for this to even happen once.)

The excerpt also contains some sensible conclusions, for instance that the reform "marginally increases the probability that a given election result might deliver control of the Senate to the Coalition".  What the report omits to mention is that it substantially increases the probability that a given (better) result might deliver control of the Senate to Labor and the Greens.  On this basis it then unfortunately goes on to conclude:

"The JSCEM model heightens the risk of a Government Senate majority and therefore doesn't achieve objective 1c: There is no evidence that the JSCEM model disadvantages any major parties, other than removing the chance to be elected from micro-parties"

Classic ALP defeatism.  They conclude that since the system slightly increases the Government's chances of a workable Senate that therefore it must disadvantage the Labor Party.  They don't even consider that it would have helped the Labor Party achieve a more workable Senate itself, for instance by ridding it of Steven Fielding (who was put there, by, er, the ALP).  They also seem to think there is some actual difference between a Senate completely controlled by the Coalition and one in which it falls a few seats short of a majority and governs with a leg up from Bob Day or David Leyonhjelm.

Of course the real problem is that the ALP so dreams of a Senate where it doesn't have to work with the Greens that it cannot see even an easier path to an ALP-Greens Senate as a win.

BTL reformed (Wednesday)

Excellent news: the Government has agreed to allow below-the-line voting from 1 to 12, with a savings provision for any vote with six squares consecutively numbered starting from 1.  It is unclear at this stage what will happen with the existing savings provisions for votes that number lots of squares but make an error early in the sequence (eg two 4s).

JSCEM's advisory report can be found here.  It contains two dissenting reports, by Senator Ricky Muir and the ALP.  Senator Muir's report is 34 pages long, thorough and well worth a read.  Labor's is a few pages long, flimsy, shallow and political and in places (eg its verballing of Antony Green over whether he said the Coalition would win 38 seats) intellectually dishonest.

Legislation is now in the process of passing through the Senate.

Savings Provision Oddity (Thursday)

The Government has now released its amendments and as I suspected, the savings provisions for below-the-line votes do require at least six squares to be correctly numbered, starting with 1.  The odd thing here is that some votes that are saved under the current provision won't be saved under the new one, so it will be very important for those voting below-the-line to make sure they number the squares 1 to 6 once and once only.  We may see some increase in the below-the-line informal rate as a result of omissions and duplications, especially from BTL voters who try to vote from the bottom up and get confused.  Generally these sorts of issues are exposed the first time a new system is tried and then can be fixed for later elections.


One more mistake has been spotted in my simulations: for the 2007 Victorian Senate for a double dissolution seats should be 6 Labor, 5 Liberal 1 Green, not 1 Family First as per the table.  This makes the correct totals for the 2007 double dissolution 33 Coalition, 33 Labor, 7 Green and 2 NXG.


  1. Hi Kevin,

    I haven't checked the numbers, but some comments seem to be inaccurate for the Double Dissolution results given.

    2013: L-NP needs 7 cross-bench votes, comment should be "GRN, 7/12 others can each pass LNP bill"

    2001: L-NP needs 5 cross-bench votes, comment should be "GRN, DEM can each pass L-NP bill"

    1998: L-NP needs 7 cross-bench votes, comment should be "DEM, GRN+PHO can each pass L-NP bill"

    1. Thanks very much for those. I'll correct the official version if I can do so in time!

    2. Corrected versions now posted.

  2. seems that it is unclear what will happen under the new system.....but why the rush, aim should be to maximise voter choice
    and minimise informal votes. That includes allowing people to vote a party ticket if they so wish. SAvings provisions should leave a vote formal to the extent that it can be counted.

  3. Above The Line Voters no Longer have choice of Voting Fully Preferentially and Preferencing ungrouped independents who will be disenfranchised from #ATL voters

  4. This is true, but it is also rather irrelevant in my view, given that 1.) voting for ungrouped independents below the line will become so much easier than it previously was, and 2.) ungrouped candidates often fared extremely poorly on ATL preferences anyway - there was no point doing preference deals with them since they had no ATL preferences to deliver, and 3.) even if ungrouped independents had done well on ATL preferences under the old system it would have done them no good since they would have been eliminated too early in the count to benefit.

    The new system will greatly improved the lot of ungrouped candidates overall even if ATL votes cannot reach them. They had no chance whatsoever under the old system but now if they can convince enough voters to just vote 1 to 12 (compared to 1 to 6 for parties) they can be elected. True there will be some loss to informal but it is trivial compared to the improvement of only needing to ask voters to vote 1 to 12 and not 1 to several dozen (with the alternative of just voting 1).