Wednesday, December 25, 2013

Legislative Council Voting Patterns Revised (2010-2013)

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 since the last Lower House election in 2010.

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

3. Excluding Tony Mulder (who is unpredictable) and Adriana Taylor (typically left on social issues, right on "moral issues"), the remaining MLCs can be considered to lean to the right to varying degrees.

4.Despite this, there is a great diversity of positions on the "right" side of the Council and a lack of strong clustering in voting patterns there.

5. Only one MLC, Rob Valentine, occupies a position to the left of the sole endorsed Labor representative.

6. As a result of the high percentage of contentious high-profile issues in the last year, differences between the various "centre-right" MLCs and Liberal Party positions have become clearer.

7. The result is that the revised analysis now shows most of the conservative MLCs to be between Liberal and Labor positions, albeit closer to the Liberals, rather than to the right of the Liberals as appeared to be the case based on the more limited data available last year.

8. The perception that the Council is a plaything of the Liberal Party on major issues generally is not supported by this analysis.  Indeed Liberal positions have been defeated on most divisions in the past year.

Correction: This article as first published described Mike Gaffney as "Independent Labor".  This designation was incorrect and has been removed.  Although having stood as a Labor candidate for the Lower House in 2002, Mr Gaffney resigned from the party in 2009 prior to contesting the seat.  Although he was seen as an "independent Labor" candidate at the time (including being recommended by then-Premier David Bartlett) he is not known to me to have employed independent Labor branding or to have had any association with the party since his election. 

It's been an unusually busy and contentious year for the Tasmanian Legislative Council, though some may say that that is not saying a lot.  There were almost as many divisions on the floor of the LegCo this year as in the previous two and three-quarter years combined.  Now that the Council is in recess for summer (and because I just like posting articles on Christmas Day) it seemed like a good time to revisit a piece from last November, Legislative Council Voting Patterns since the last Lower House election.

The previous article was based on a sample of just 25 divisions and some of the conclusions were a little bit surprising.  For instance it placed Kerry Finch, usually thought of as a leftie by LegCo standards, as centrist or even very slightly leaning to the LegCo right.  It also placed Rosemary Armitage well to the right compared to her reputation.  With a much beefier sample of 51 distinct divisions, it's time to see if the initial assessments are still valid, or if the small sample of bills covered in the previous article resulted in the odd unrepresentative result.

The 26 divisions added here include some very controversial items - forestry (8 distinct divisions, 2 of those in late 2012), Anti-Discrimination law (4 divisions), abortion law reform (3 divisions), same-sex marriage (1 division) and even the remarkable appearance of a division on whether to read a bill the first time.  Five MLCs were so chagrined by the government's late introduction of electoral spending changes that they didn't want the bill to pass first base (its end fate was to be stranded on the notice paper at year's end, killing any chance it would apply to the 2014 state election).  There were some obscure items but it was a big change from the time leading up to my previous article, in which most of the contested divisions had been on low-profile issues.

This analysis deals with the form of the current Legislative Councillors since the 2010 state election, and includes all the divisions in that time, including those analysed in the previous article.  I include figures for Jim Wilkinson, although his elevation to the presidency means that he has not voted since April, except where required to exercise a casting vote (in which case there are established conventions that can be expected to override his own opinion.)

The sample includes one rather funny case.  A motion to remove the automatic veto on cable cars on Mt Wellington (meaning that cable car proposals could be considered under normal development rules) was strongly supported by the Council, but opposed by Rob Valentine MLC.  A division was called for and it soon turned out that all the MLCs except Valentine were in favour.  Jim Wilkinson admitted to being in somewhat of a quandary about who to appoint as teller for the Noes. Upon the 13-1 result being formally declared, Valentine called out "recount!"

I've again aimed to produce two representations of the voting patterns.  One is two-dimensional and the other is a left-right sort.  Again this closely follows the methods of an old TT piece Your City Council Sorted From Green To Blue 2009-2011, with the various exceptions discussed in that methods piece not needed.  Because Wilkinson has been President since Hiscutt's election, I had to calculate an estimate of how often Wilkinson and Hiscutt would have agreed had they been voting members of the same council.  The estimate I came up on, based on a series of linear regressions involving other MLCs, was 64%.

In Two Dimensions

The following is a two-dimensional representation of the voting patterns of the fifteen current MLCs since the 2010 state election (or since they were elected if they were elected since).  As discussed above Jim Wilkinson no longer votes on the floor of the LegCo but is included anyway (it's possible he will return to doing so in future if he does not wish to serve a full term in the chair.)

For those unfamiliar with graphs of these sorts, a principle 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 indicate what kind of voting pattern is displayed and the distance indicates how strongly it is realised.  Even if two Legislative Councillors appear opposite each other, if one is close to the centre they will still agree reasonably 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.

Key to names: Arm = Rosemary Armitage, Dean = Ivan Dean, Far = Craig Farrell (Labor), Fin = Kerry Finch, For = Ruth Forrest, Gaf = Mike Gaffney, Good = Vanessa Goodwin (Liberal), Hall = Greg Hall, Har = Paul Harriss, Hisc = Leonie Hiscutt (Liberal), Mul = Tony Mulder, Ratt = Tania Rattray, Tay = Adriana Taylor, Val = Rob Valentine, Wil = Jim Wilkinson

(Although twelve MLCs are "independents", some of these have known past or present party connections.  Gaffney is a past Labor candidate. Mulder is a recent past Liberal candidate who was elected as an "independent liberal". Taylor is a former Labor Party member.  Harriss was a Liberal candidate a very long time ago and is seeking preselection for the Liberal Party for Franklin next year.  Valentine was once a Green Independent candidate (but not a party member) decades ago.)

In comparison to the 2010-12 graph  there are several differences.  The left side of the graph is more compressed with Finch clearly joining it.  Hall, Harriss and Wilkinson remain quite strongly clustered, but Taylor has moved to a position between them and the left, Armitage is indicated as only mildly conservative and Rattray has moved from a position near Dean to one closer to the main cluster.  Goodwin, meanwhile, is no longer really part of the cluster, and the other Liberal, Leonie Hiscutt, is well away from it. 

My reading of what's going on here is that the LegCo is showing itself to be not quite the Liberal Party playpen that some think.  That is indeed supported if we look at the percentage of times each member has been on the winning side of contested motions they have either voted on or been paired for on a specific side:

Hall 78, Wilkinson 77, Finch 76, Gaffney 75, Taylor 73, Armitage 72, Farrell 69, Forrest 66, Harriss 63, Rattray 63, Goodwin 60, Mulder 57, Valentine 56, Dean 51, Hiscutt 33

Compared to the previous article, Farrell's win rate is up nine points, while new Liberal Hiscutt has a much lower strike rate than anyone did in the 2010-2 period.  As Goodwin voted with Hiscutt on all but two motions since Hiscutt's election, Goodwin has now been on the losing side 40% of the time in this term (up 15 points). 

Some may say that the Liberal Party is winning all the big points and the left is only scoring on the margins.  I don't agree.  For instance, the forests peace deal, passed with Liberal opposition, is now strongly supported and praised by much of the moderate left (including some of those who originally considered it to have been gutted or vandalised by Upper House amendments.)  

Left to right sort

As the previous article notes I do find that left to right sorting tends to compress information and oversimplify, but can also provide useful new information.  The following is an agreement matrix, which shows the percentage of divisions on which the two MLCs, if both present (or paired for voting purposes as on a specific side), voted on the same side.  For instance the table shows that Mike Gaffney and Kerry Finch voted together 80% of the time.  The highest agreement percentages were 93 (Farrell-Gaffney), 89 (Hiscutt-Goodwin, both Liberals) and 88 (Valentine-Forrest).  

The lowest agreement percentages were 6% for Valentine-Hiscutt, 15 for Valentine-Dean and 17 for Forrest-Hiscutt.

A figure of three-quarters (75%) is often used as a threshhold for identifying clusters of voters with similar views in this sort of work.  Therefore I have highlighted agreement scores of 75 and above, and also weakly highlighted those of at least 70.

I sorted the MLCs from left to right (by the standards of the Council, anyway!) using the same comparative sorting method as last time (details as per the Council methods piece, which casual and general readers should avoid).  I also used my older Council assessment method of computing ratios of agreement with a fixed list of "left" and "right" (treating Taylor and Mulder as the centre).  The two methods produced an identical order of MLCs but this time I used the second method to produce ratio scores for each MLC.  I did this because the fragmented nature of the right of the Council resulted in the first method deeming all of Taylor, Mulder and Armitage to be on the council's left, and I think that's clearly false in the case of Armitage at least.  Note therefore that the ratios are not on the same scale as last time, and haven't been scaled.  Ratios on red or orange represent left of centre and any shade of blue means right of centre:

The cluster of five left MLCs in the top left hand corner is very clearly defined and obvious.  On the right side things are messier with weaker agreement percentages in most cases and a concentration of 60+% agreements, but not clearly defined clusters featuring very strong agreement between all those included.  As well as some right-wing MLCs being more extreme than others, there seem to be a few different brands of right-wing views represented.   The sorting tends to confirm the impression from the graph above that most MLCs are actually between the Liberal and the Labor party lines, once enough big-ticket items are voted on.

Changes compared to the previous article are:

* Forrest is now rated slightly to the left of Gaffney.
* Armitage is now rated as among the most moderate of the right-leaning MLCs, suggesting that her rating as the second most right-wing in the previous article was probably a result of the limited sample of mostly obscure issues canvassed during the start of her term.
* Hall is now shown as more moderate than Harriss and Rattray, though in the first case by very little.
* Because of following the party line on a number of lost votes, Goodwin is now shown as third most right-wing.
* New MLC Hiscutt is shown as the most right-wing.  That may partly reflect the increased percentage of party-line issues since she was elected, but it also partly reflects her social conservatism, and probably would have happened anyway. It's my view that Hiscutt probably is the Council's most right-wing member under normal understanding of the term. 

In this case, the left-to-right sort and the 2D picture tell a very similar story.  In the coming year, fifth on the left Kerry Finch will face a challenge from the Liberal Party and Paul Harriss will throw in the towel if endorsed for the state election (see article re Harriss here).  Peter Hodgman (an ex-incumbent of the seat from way back and a former Liberal MHA and Minister) is running for Harriss's seat of Huon.

OK, that's probably enough pseph for one Christmas Day!

(PS re the lack of articles for the past nearly two weeks, I spent much of that time overseas.  Normal service should now resume for a while, whatever that is.)


  1. Just to clarify, the 2D chart is a generated difference plot, rather than a Nolan chart? I suspect the two may map pretty well to each other, even if generated by different methods.

  2. Correct. It's just fortunate that the PCA graph has put the lefties on the left and the righties on the right; there was no deliberate prodding in that direction and the PCA is not told that issues are of any particular sort. The generated x-axis appears to correspond extremely well to "left-vs-right" as generally perceived, but to a multi-issue understanding of it rather than a single dimension like economic or social issues.

    I do not think the generated y-axis corresponds to anything that is easy to define. For instance a major difference between the main conservative cluster on the one hand and Hiscutt and Dean on the other is that the MCC are less socially conservative, but Taylor is more socially conservative than most of the conservatives, and Mulder less so. Those two at least would be somewhere different if it was a conventional Nolan chart. Both the 1D and the 2D assessments make Taylor look less politically distinctive in this setting than she is.