Sunday, April 2, 2017

Legislative Council Voting Patterns 2013-7

(Note: for updates on the Braddon recount go here)

Advance Summary:

1. This article presents a revised analysis of voting patterns in the Legislative Council (the upper house of Tasmanian Parliament) based on contested divisions in the last four years.

2. Although there is a degree of independence in all Legislative Council voting, the Council continues to have a clearly defined "left wing" consisting of Craig Farrell and Josh Willie (Labor), and independents Mike Gaffney, Ruth Forrest, Kerry Finch and Rob Valentine.

3. Excepting Rosemary Armitage and Tania Rattray (and Jim Wilkinson, who does not vote) the remaining MLCs (independents Ivan Dean, Robert Armstrong, Greg Hall, independent Liberal Tony Mulder and endorsed Liberals Vanessa Goodwin and Leonie Hiscutt) can all be clearly placed on the "right wing" side.

4. A possible left-to-right sort of the Council could be Valentine, Forrest, Gaffney, Farrell and Willie, Finch, Armitage, Rattray, Hall, Armstrong, Dean, Goodwin, Mulder, Hiscutt.  However most of the exact positions in this list are debatable.

5. Voting in the Legislative Council was again not very party-polarised in 2016.

6. The Legislative Council is finely balanced going into the 2017 elections.


We're gearing up for another season of Tasmanian Legislative Council fun and games with three incumbents facing the music in May, and eight hopefuls known to be challenging them already.  With the Council now finely balanced, MLCs Forrest, Armitage and Mulder will try to defend their positions, after last year's ousting of former Glenorchy Mayor Adriana Taylor showed that the red couches are not quite the safe seats that they used to be.  As my traditional (and traditionally super-wonky) curtain-raiser for my coverage of the contest, here's my annual review of the voting patterns displayed by the current MLCs in the last four years.

For previous articles on this see last year's piece, which in turn links back to previous years.

Every time I do this review, I only include the last four years of data, which means no Jim Wilkinson this year as it's been four years since his last non-casting vote. I only include votes where there were at least two MLCs on each side of the floor, and not the many unanimous votes or cases where no division was called.  Since the last review there have been another twenty votes to include.  Issues covered in these have included politicians' pay, Uber licences, anti-discrimination law, education, community protection bills and gaming.  There have also been symbolic motions on same-sex marriage and Aboriginal names for electorates.  In general, there have again not been that many big-ticket left-right stoushes coming upstairs from the Lower House, and so there have been more chances for MLCs to display independence from each other.  Or not, as the case may be.

The Council even found time to have a split vote on an amendment to an amendment (the principle issue being a gaming enquiry).  I've been on the Australian Chess Federation Council for 17 years and as weird as chess politics is, we have never had one of those!

I've again aimed to produce a couple of different descriptions of the observed voting patterns.  One of these is a two-dimensional graph and the other is a left-right sort.  Again this mainly follows the methods spelt out in the ultrawonky PDF attachment to an old Tasmanian Times Hobart Council article.

A number of judgement calls were made in the left-right ordering, which is much more rubbery this year than normal, and these are discussed later.  By the way I often get requests, invariably from the left, to separate procedural and substantive motions, but I don't do it.  Split motions that are clearly procedural in the LegCo are rather rare, but in my experience motions that appear to be procedural can sometimes be proxy for something more substantive anyway.

In Two Dimensions

The following is a two-dimensional view of the voting patterns of the fifteen current MLCs over the last four years.   For those unfamiliar with graphs of these sorts, a principal components analysis aims to represent patterns in 2D with as little distortion as possible.  Both the angle of different lines to each other and the distance of different data points from the centre are relevant here.  The angles between different candidates indicate whether or not they display different kinds of voting patterns and the distance indicates how strongly each pattern is realised.  Even if two Legislative Councillors appear opposite each other, if one is close to the centre they will still agree fairly often.  If two Legislative Councillors are at a similar angle and a similar distance from the centre then it is likely their political views are rather similar. The two axes chosen by the analysis do not necessarily mean anything in particular and are not predetermined by me, but it's obvious in this case that the x-axis corresponds pretty closely to "left-right".

The "left-right" axis explains 75% of all the variation in the extent to which different MLCs do or don't agree with each other.  The other axis only explains 9% and might be called the Rattray factor.  In policy terms, it doesn't have any clear definition, and it's possible that the left-right axis is the only real pattern there is and the second axis is mostly noise (see comments.) If I went to three dimensions, a third axis explaining just less than the second one would have Rattray closer to the centre but Armitage well away from it, so the pattern that these two are basically near the centre (but in different ways) while everyone else can be placed on the left or right is well supported.

(Arm: Rosemary Armitage (Ind), Arms: Robert Armstrong (Ind), Dean: Ivan Dean (Ind), Farr: Craig Farrell (ALP), Fin: Kerry Finch (Ind), Forr: Ruth Forrest (Ind), Gaff: Mike Gaffney (Ind), Good: Vanessa Goodwin (Lib), Mul: Tony Mulder (Ind Lib), Hall: Greg Hall (Ind), Hisc: Leonie Hiscutt (Lib), Ratt: Tania Rattray (Ind), Val: Rob Valentine (Ind), Will: Josh Willie (ALP))

The placement of Josh Willie on the chart is a bit unreliable because only sixteen votes by him were included.  Gaffney and Finch appear together not because they always vote together (they don't) but because the differences in how often they vote with other MLCs are basically random.  Goodwin and Hiscutt, on the other hand, have almost always voted together, an exception this year being a vote to support same-sex marriage.

The main change this year is that Rattray is even further away from the right-wing cluster, while Valentine is slightly closer to the rest of the left.  These changes are minor as the removal of Taylor and Wilkinson from the sample could well affect how the PCA sees things.  (The fact that the graph is upside-down as concerns the position of Rattray compared to last year's is irrelevant.)

Left-right sort

As usual the agreement matrix below shows some similar patterns to the PCA graph.  The matrix shows the percentage of contested divisions on which each pair of MLCs voted together.  For instance it shows that Goodwin and Armitage have voted together 57% of the time.

The highest agreement percentages are 100% for the two Labor MLCs (I've assumed the order of names was inverted in the Pair listing for one motion, which otherwise seems to show Willie on the other side to Farrell), 97% for the two Liberals, 90% for Forrest-Gaffney and 88% for Hall-Armstrong.  The lowest are 23% for Valentine-Dean and 24% for Valentine-Hiscutt and Farrell-Mulder.

As usual I've highlighted agreement percentages of 75+% (a common cutoff for identifying clusters) and weakly highlighted those between 70 and 74.

As in last year, I give two different alignment scores.  Score1 reflects how strongly the MLC tends to vote with the left of the Council (red) or the right of the Council (blue), rather than the other way around.  Rattray and Armitage are considered neither right nor left, although both are very slightly closer to the right.  Score2 is based on ratios between the MLCs based on whether a pair of MLCs are more likely to vote with those to the right of them than those to the left (again this is explained in the ultrawonky HCC methods piece.)  For the purposes of Score1, agreement scores involving Willie were downweighted by a factor of three because of small sample size.

* Note: Limited data for Josh Willie so treat all numbers with caution, see comments below

There is an obvious cluster of high scores in each corner, and a couple of MLCs who don't have any 70+% scores with anyone.  These are the same left-right clusters we have seen before but now there are six on the left and six on the right.  Not only did the left gain a seat last year but also over time Tania Rattray, once one of the more conservative independents, has moved towards the centre.

The small sample size for Josh Willie and the relatively non-partisan issues mix means that all the figures involving him have to be treated with caution.  One would not expect that he and Mike Gaffney will only vote together half the time in the long term, assuming Gaffney's normal voting pattern continues.

There are a lot more judgement calls in my ordering of the MLCs this year than normal:

Valentine and Forrest: Score1 (from left to right) says Forrest-Valentine, Score2 says Valentine-Forrest.  Inspection of the figures shows Valentine is less likely to agree with MLCs on both the left and the right.  Score2 is good at picking this sort of thing up if the reason is that someone is more to one side then the rest of their cluster, so I've gone with Score2 and kept last year's ranking.

Gaffney, Farrell, Willie and Finch: The problem here is we have four years of data for Farrell but only one for Willie, and although the two have always voted together so far and possibly always will, the Labor position falls closer to the centre in the last year than normal.  So things get messy - Score1 says Farrell-Gaffney-Finch-Willie and Score2 says Gaffney-Farrell-Finch-Willie.  It makes no sense to separate the two Labor MLCs at this stage and the evidence from Farrell's track record is likely to be more useful.  So I've gone with Gaffney-Farrell-Willie-Finch.

Armitage and Rattray: Score1 says Armitage-Rattray by a small margin, score 2 says Rattray-Armitage at the fourth decimal place.  There's no reason Score2 should be better in the middle of the pack so I've gone with Score1.

Dean, Mulder and Goodwin: Partly because of Goodwin's high agreement score with Hiscutt although Goodwin is more moderate on social issues, Score2 suggests Dean-Mulder-Goodwin and Score1 suggests Goodwin-Dean-Mulder.  I've therefore gone with Dean-Goodwin-Mulder.

We are seeing generally that there is more distance between the Liberals and the "independent Liberal" Mulder on the one hand and the conservative independents, with more data for Armstrong showing that he is less conservative than he initially seemed, and Hall also now closer to the middle than he was.

In the last year, the government got its way on slightly more than half the actually meaningful motions, and one of the losses was the attempt to disallow a pay rise for politicians (the sort of motion they may have secretly wanted to pass, so long as they were seen on the losing side.)  That one aside, none of the losses were remotely front-page stuff, but that could change.

This fine balance of power makes for an election that could mean everything or nothing.  Those facing election are one from the left (Forrest), one from the centre (Armitage) and one from the right (Mulder).  If the left gains another seat then the Hodgman Government could find it very hard to get anything really contentious through the chamber in the last year of its term.  Federally, you'd start looking for a double dissolution, but Tasmania doesn't have those, so the government would just have to deal with it - including if it won the next election.  The Liberal Governments of old, in the terrible days before one-vote-one-value, didn't have to worry much about this problem!

Legislative Council guides should be posted here this week.


Addendum (5 April): I'm commenting on a comment about this article posted by Brent Smedley on Ruth Forrest's Facebook page, following an attack on Forrest's voting record by her opponent.  Brent writes:

"I read Kevin's blog not long after he posted it and my first thought was that his labelling of voting as left or right had the potential to cause misunderstandings. There are several disclaimers in the blog that many votes in the LegCo are procedural. Another way to interpret his statistics are that people on the right are more in lock step with the government while those whose voting record is on the left are more independent. Traditionally independence has been seen as a positive trait for a legislative councillor. Unfortunately your typical voter isn't going to get that distinction."

The problem with this objection, as I read it, is that it's empirically incorrect.  Those who are on the right were by and large also on the right when Labor was in government - they usually voted against Labor and with the Liberals then and are doing the same thing now.  Those who are on the left were also on the left when Labor was in government - they usually voted with Labor and against the Liberals then and are also doing the same thing now.  The only ones who have moved all that much are Rattray who has moved from right to centre, and Mulder who has done the opposite.  It might be suggested from this that Rattray has been cautious about both governments while Mulder has some tendency to try to give the government of the day a go.  Obviously given his Liberal sympathies that would be more pronounced with the Liberals in power.

A consistent left or right position says nothing about whether people are more or less independent of governments of the day than the other side.  It says only that lefties will be more independent of right-wing governments and righties will be more independent of left-wing governments.  That's hardly news to anyone.


  1. After people have run this correlation-based sort of analysis which generates axes without prior reference to ideologies, it's usually possible to look at the issues that they've voted on and give some sort of name to the axes. Here you've noted that the main axis is pretty clearly the usual left-right stuff. So is there any sort of label that you can give to the "Rattray factor", or to the "Armitage factor" that seems to form your 3rd axis? Do the issues where they vote one way or t'other, that generate these axes in your computer, seem to be "social" isues, or local issues, or does it all just seem idiosyncratic and beyond classification?

    1. My suspicion - the more I've thought about it - is that in this year's PCA graph specifically, the Rattray and Armitage factors are artificial and result mainly from the pre-setting of a high agreement factor for each MLC with themselves. If this is right then for this year specifically, everything is really noise except the left-right pattern, which displays itself across a wide range of issues (especially forestry and culture-war issues like anti-discrimination law).

      This hasn't always been the case. For instance for the 2010-3 analysis ( the 2D graph distinguishes between the Liberals and most of the conservative independents, so the second axis in view of Mulder's voting at the time looks something like ruralism vs libertarianism (but slightly tilted).