Monday, December 8, 2014

Fisher, State By-Elections and Federal Drag

Nat Cook (ALP) provisionally won, defeating Dan Woodyatt (Ind) by 226 votes at the critical exclusion point then Heidi Harris (Lib) by 23 votes.  This is currently being recounted, with Cook defeating Woodyatt by 219 in the recount and defeating Harris by nine votes. Cook will now take her seat and Labor will have an outright majority.

The win may well be challenged in the Court of Disputed Returns.  In this case Labor will hold the seat until such time as the court decides the case.  The court could confirm the win, order a by-election or declare the Liberals to have won the seat.  

Postcount Introduction

This article follows late counting in the South Australian Fisher by-election.  As at the close of counting on the night this seemed like a very likely win for the ALP who had a 52.1% two-party preferred lead against the Liberals.  Even with a very large percentage of prepolls such leads are very rarely caught.  Although there was still some threat from the independent Dan Woodyatt, I thought I should place the at that stage apparently likely ALP win in its historic context.  That was the theme of the original article. 

Events since increased the chance that the light shed on the history of state governments gaining seats at by-elections would be only a footnote.  Although the Liberals suffered a severe 2PP swing, their candidate Heidi Harris came storming back into contention on postcount votes, but is currently trailing very slightly.  A by-election is an easier environment for a Liberal to stage a post-count recovery in than a booth election since there are no on-the-day absent votes (which favour Labor).  Even so and even with the size of the post-count the scale of recovery seen here is a rare one.   It seems to have happened not only because of changes in primary votes since the day (which were pretty much as expected) but also because the distributions of preferences from specific candidates were very different in the pre-polling than on the day.

But it is not only Harris who came back into contention but also Woodyatt.  Normally independents struggle in prepoll and postal voting slightly, but in this case he made potentially crucial gains and became a much more serious threat to Labor than he was on the night.

The following is my current assessment (3:30 pm Monday during recount) of the key questions:

1. Will Cook (ALP) or Woodyatt (IND) finish second?   Cook won by 219 votes after recount (previously 226).  Prior to that my assessment was as follows: Woodyatt will be third on primaries, but the question is how much of the gap he can close to Cook before the distribution of the preferences of minor candidates (one of whom preferenced him, the rest preferencing nobody).  The figure I am tracking here is how many points more of the minor candidate vote he will need compared to Cook.  For instance if Woodyatt gets 55% of the minor candidate preferences, Cook 30% and Harris 15%, then his gain is 25 points (55 minus 30).  Woodyatt's target started at 36.7% and has come down to 24.5% on current figures.  In the absence of clear scrutineering information, he would probably want it to come down just a little more. 

2. If Cook is second, who wins between Cook and Harris?  The Liberals were provisionally eliminated from this race by 23 votes.  During a recount this margin reduced to nine.

3. If Woodyatt is second, who wins between Cook and Woodyatt?  I have seen no scrutineering on this but expect Woodyatt would win easily on Labor preferences against a Liberal, from any set of primaries that might be reached from here.

Postcount and recount updates scroll below the line with the most recent at the top.


Monday 11:30: William Bowe notes: "The recounted results have moved 16 votes from formal to informal, reducing the Labor count by seven, the Liberals by three, Woodyatt and the Greens by two, and De Jonge and Couch by one. On 2PP, Labor lost 15 and the Liberals lost one." These are net changes as there would have been movement of votes in both directions.

Monday 10 pm: Nine votes! So Cook will be officially confirmed as winner after all, but there may yet be a court challenge.

Monday 8:40pm: Tom Richardson and others report that following checking of all disputed votes the margin is now 13 and this is being checked during a final check of the preference distribution.

Monday 4pm: Announcement of the recount result is "hours away" according to ECSA via Michael Ramsey.

Monday: ECSA is conducting a recount of the seat at the request of the Liberal Party.  The votes have been subjected to a large level of checking already but arguments are still being made concerning the formality of various contested votes.  The recount result is expected late this afternoon.  Tom Richardson reported at c. 2pm that the Liberals had clawed back three votes to make the margin 20 with nine votes previously declared informal being overruled.  Richardson also noted "Sorted all informal votes & decided 9 now formal. Sorted all Cook/Harris & raised Qs on validity of several, to be adjudicated."  It is less likely (on a case by case basis) that votes previously ruled formal will be ruled informal than the other way around, or that many will still be in the wrong pile, but there are far more formal votes than informal so we will need to wait for the final totals. 

Postcount Updates

Saturday 2:20 I've been playing in a chess tournament today but it's being reported that Cook has beaten Woodyatt at their exclusion point by 226 votes and is therefore the winner!  All that was reported in the original article about the historic nature of Cook's win applies (NB subject to recount) and the margin is irrelevant.

Friday 6:40: It's being widely reported that following thorough rechecking the Liberals have lost the 2PP race by 24 and are out of it.  See, I told you they were too far back. :)  I'm on the road so cannot post further at the moment.

Friday 2:00: Thirty postal votes were added today breaking 15-15 on a 2PP basis.  Woodyatt's gain rate required over Labor went back up to 24.7%.  Tom Richardson also reports that vote-checking has so far added two votes to Labor's position - realistically Harris now needs a counting error or uncounted vote discovery in her favour, otherwise tomorrow is Cook vs Woodyatt on minor candidate preferences to make the final two, and whichever of them does so wins.

Thursday 1:00: Nine postal votes were received today.  The 100 or so outstanding are those that were issued that have not yet been received; probably many of them weren't mailed back with a postmark by election day, so there may not be much more to add at all.

Wednesday 5:00: William Bowe advises that there are only the few remaining postals (believed to be no more than 100) remaining and that they'll be done on Friday.  On Saturday I will be playing in a chess tournament and offline for hours at a time (at least) so suggest the Poll Bludger thread as a place to keep an eye on the outcome.  Or for those on Twitter, #fishervotes and @DanWillsTiser.

Wednesday 12:00:  Well, normally postals favour conservatives but apparently in this unpredictable by-election late postals don't (probably for the same reasons as drove the discrepancies between booth voting and prepolls - or perhaps just because the number of votes remaining is so small that they are less predictable).  Postals today split 52-52 for the lead candidates with 47 for Woodyatt and on preferences Cook was back to a 21-vote lead.  If Cook-Harris is the key contest it will probably be recounted, but Woodyatt's required rate of gain on minor candidate preferences came down again slightly to 24.5%.

Tuesday 4:10 pm: It's apparently turned out that the flow from the prepoll and postal preferences to the Liberals (43.3%) was much more favourable than in booth votes (34.5%), and this, coupled with the strength of the prepolls, gives them a lead of 17 votes, which will probably increase as postals are added, though Labor may also have chances on the provisionals.  The logic behind this (and the error in my assumptions above) is too obvious in retrospect: as well as the smaller proportion of tiddler preferences coming from the Greens, the Liberals would have cost themselves Woodyatt preferences when they attacked him late in the campaign.  Therefore their share of Woodyatt preferences in prepolls would be higher.

Tuesday 3:30:  A large number of prepolls have been added to the primary count.  Overall Harris (Lib) is now on 36.1% compared to 35.0% on the night, Cook (ALP) 26.7% compared to 27.9% and Woodyatt (Ind) 23.3% compared to 22.5%.  The Greens have dropped from 3.9% to 3.4%, others have moved little.  I'd expect from that that the Liberals would pull the 2PP back to something around 50.6, which wouldn't be enough.  Woodyatt is an increasing threat, however, now needing to beat Labor on minor candidate preferences by 24.7 points, which is much more plausible than the previous asking rate.

Tuesday 1:23: A reported sample of prepolls had a near tie, 154-153 to Harris (Lib) on a 2PP basis (she needs about a 55:45 split).  No way of knowing if these are representative so I'd wait for more. Reported by David Washington (InDaily).

Tuesday 12:30 pm: Tweet from Daniel Wills (Advertiser) indicates that based on a small sample of prepolls, Labor does not seem to be doing worse compared to booth votes.  Most of the remaining votes should be counted today and if the Liberals do not make big 2PP gains on the prepolls today then it's all over for them.  The Woodyatt question might remain technically unresolved until the throw of preferences on Saturday.

Monday 1pm: The addition of postal votes has significantly assisted the Liberals as usual, breaking 57:43 to them and cutting Labor's 2PP back to 51.43%.  It is normal for postal votes to heavily favour the conservatives and remaining declaration votes probably won't help them as much; even if they do close the gap significantly Labor is still very likely to get home.  More importantly, Woodyatt stayed in the game for now by making some gains on Labor on postals.  On current primaries the rate at which he has to beat Labor on minor-candidate preferences (in a three-way throw to Independent, Liberal and Labor) has dropped from 36.9 points to 32.2.  It may well increase again when prepolls are included.

Original Article

 Advance Summary

1. On current figures the ALP has probably won the Fisher by-election in South Australia, and hence a one-seat majority.  The Liberal Party has suffered a large two-party swing and has almost certainly failed to win the seat.  An Independent is still competitive in the race for the seat.

2. A Labor win was almost universally unexpected. 

3.  Cases of state governments gaining seats at by-elections are rare events, but there have been at least 31 such cases since Federation.

4. State governments are more likely to gain seats in by-elections when they are in their first term in office and when their party is in opposition federally.

5. The Fisher case, if confirmed, will be the first case of a by-election gain by an incumbent government in the 78-year history of single-seat state elections in SA, and the first case in any state involving a non-first-term government since 1973.

6. The unpopularity of the Abbott federal government is a major factor in the 2PP swing to Labor, but is not alone sufficient to explain it.

7. A variety of state and federal factors are likely to have contributed to this surprise swing and potential upset.


With the dust still settling from the Victorian state election result, in which federal factors were widely and correctly considered to have played a major role (see my Victorian election wrap for more on this) another electoral event has again brought the question of federal impacts on state elections into the spotlight.  The Victorian election was an utterly predictable result, but this one is being seen as a big surprise.

The ALP appears to be winning the by-election for the South Australian seat of Fisher, vacated as a result of the death of sitting member Bob Such.  Such had held the seat for 25 years - as a Liberal from 1989 to 2000 and as an independent thereafter.  In this time the Liberals had beaten Labor in the two-party preferred vote in the seat six times out of seven, most recently 57:43.  The seat was widely expected to either revert to the Liberal Party or else to fall to independent Dan Woodyatt, an ex-ALP member endorsed by Such's widow.

While the seat's Liberal-friendliness had been exaggerated by a series of elections at which Labor had no chance of winning the seat from Such hence probably didn't try hard, practically nobody fancied the chances of the 12-year-old state Labor government.  They had good reason not to, given that said government flopped across the line in minority in this year's state election after losing the 2PP rather heavily (47:53).

Even popular governments can have a hard time holding close seats at by-elections, let alone picking up seats with a 2PP swing currently sitting at 9.4%.  So just how common would such a result be?

(Disclaimer: the seat's fate is not certain yet.  While Labor's 2PP lead over the Liberals (52.1% 2PP at the time of writing after an ECSA correction) will very likely stand up to even such a large post-count vote, there is still some challenge in theory from Woodyatt.  On current figures, Woodyatt needs his percentage share of preferences from five minor candidates to be 36.9 points higher than Labor's.  So, for instance, if minor candidate preferences split 65% to Woodyatt, 25% Labor, 10% Liberal, he'd just win.  My understanding is the flow to Woodyatt is nowhere near that good, but maybe if he could close the 5.4% gap to Labor to below 3% in post-counting he'd have a hope.  Difficult though and scrutineers may already know better concerning the preferences of the tiddlers.)

Government Seat Gains At State By-Elections

Antony Green posted a comment on the scarcity of seat gains by state governments at by-elections following this result, and listed several from memory (Earlwood 1978, Clarence 1996, Burwood 1999, Benalla 2000, Mulgrave 1998, Northern Tablelands 2013, Helena 1994).  Antony noticed that in all these cases (some being gains from the opposition, some from independents or minor parties) the state government was in its first term.  Not only is a first-term state government often popular, but also it can benefit from the retirement of defeated heavies from the previous government.  These retirees (Kennett in Burwood a prime example) not only take huge personal votes with them, but also annoy the electorate by dragging it back to the polls. 

I, of course, noticed something else: in all Antony's cases, as in Fisher, the state government was of the opposite party to the federal government at the time. 

So I thought I'd whip up a contingency table: numbers of by-election seat gains by incumbent state governments, by whether they were of the same party as the federal government, and whether or not they were first term.  Wins by junior coalition partners are included.

The data mostly go back to Federation, but in the case of South Australia only results from the adoption of a single-seat system in 1936 onwards are included.  For Victoria I didn't find data from 1901 to 1913. Tasmania is excluded because of the use of Hare-Clark in its lower house.  In all I found 688 by-elections (251 NSW, 121 Vic, 136 Queensland, 133 WA, 47 SA).

Here's the table.  Term of state government across the top, whether or not it is the same government as federally at the side.

In the "other" category I've included some early cases where the state government was neither the federal government nor the federal opposition as such.  For instance, when the "United Country Party" government in Victoria gained Gippsland North in 1938, it was estranged from the federal Country Party and Coalition (to the extent that that year it expelled federal MP and much later PM John McEwen.)

Some of these examples scrape the bottom of the barrel.  In the case of the Davidson (NSW) state by-election in 1992, the disgraced sitting member had become an  independent just prior to resigning, upon which his seat was regained by his former party.  But most are not like that, although a fair proportion are wins from independents.

We can see firstly that it is a rare event for a seat to be gained by a state government at a by-election at all - this happens in 4.5% of by-elections.  However, probably around half or perhaps slightly more (I've not checked the exact figure) of all state by-elections are for seats vacated by the incumbent government, so government gains probably happen in about 10% of cases in which they can.   It's not a super-rare event, but it's not common.

We can see secondly that both the patterns noted from Antony's sample hold up.  There are almost twice as many cases of governments gaining seats at by-elections if the government is a first-term government.  The exodus of former government members is an obvious factor here.

There are also over twice as many cases of governments gaining seats at by-elections if they are the same party as the federal Opposition rather than the same as the federal Government.  This is consistent with the general theory of federal-state election results drag that I explored for scheduled state elections in What Kills State Governments: Age Or Canberra?

An obvious objection to the latter is that since same-party state governments are more likely to get the boot, there shouldn't be as many cases of same-party state governments being in office when a by-election is held.  But actually, because federal governments tend to come in after the other party has been shedding state governments, they tend to start with a majority of state governments on their side.  I haven't done a finely-tuned assessment of this but based on Malcolm Farnsworth's table, there's no evidence that federal governments on average are in office in fewer than half the mainland states at a given time.  (If anything, slightly more than half.)

Unusual Examples

It's worth looking at the examples that, like Fisher if confirmed, are unusual in some way - either the state government is older than first term, or the state government is the same party as the federal government.  Fisher ticks the first box and would be the first such case since 1973.  The last case ticking the second box was in 1971.

State government not first term, state and federal governments same (4 cases)

East Perth (1943) - A very token example - the fourth-term state Labor government won the seat of an "Independent Labor" incumbent who had resigned to unsuccessfully run for federal parliament.  This was around the height of Labor's federal wartime popularity.

Albury, NSW (1946) - The second-term state Labor government won the seat of former Premier Alexander Mair who had quit to run for the Senate taking his personal vote with him. Federal Labor under Ben Chifley had just been comfortably re-elected at the time.

Maylands, WA (1951) -  The fourth-term Liberal-Country League government won the seat after the death of an aligned Independent.  At this time the first-term Menzies government had yet to become unpopular.

Maryborough, Qld (1971) -  The sixth-term Country government of Sir Joh Bjelke-Petersen won this seat (which Labor had held for over 50 years) following the death of the 18-year Labor incumbent.  New PM Billy McMahon was extremely popular at the time (yes you read that correctly) but a huge factor in the result was the Joh regime's crackdowns against supposedly anarchic protests associated with the apartheid-era South African rugby tour.

State government first-term but in government federally (4 cases)

Liverpool Plains, NSW (1911) - Labor win after loss of seat by Labor by three votes at state election was overturned and by-election ordered. 

Dubbo, NSW (1942) and Lachlan, NSW (1943) - both wartime Labor wins with federal party extemely popular.

Bathurst, NSW (1967) - Country Party win from Labor after death of 32-year ALP incumbent with federal Coalition way ahead in polling at the time.

State government post-first term, in opposition federally (5 cases, Fisher would be sixth)
Kedron, Qld (1951) - The seventh-term Labor government won this seat after the Liberal-turned-Independent incumbent (who seems to have been a very principled chap) resigned.  I don't know whether the incumbent deliberately or coincidentally timed his resignation to embarrass Labor over issues that saw their win in the seat of Bulimba voided, but in any case Labor won both by-elections.

Ashfield, NSW (1952) - The fourth-term Labor government won this seat after a 14-year Liberal incumbent resigned to take up a court position.  At the time the federal Liberal government was extremely unpopular in a backlash against the original "horror budget".

Kahibah, NSW (1957) - The sixth-term Labor government won this seat after the death of the Independent Labor incumbent.

Lismore, NSW (1959) - The Country Party had won the seat by two votes against an independent, but the win was declared void.  The Country Party made the surreal decision to endorse both the sitting member and his independent rival and the eighth-term Labor government (which hadn't even contested the seat in 1958) laughed all the way to the bank.

Murray, NSW (1973) - The 41-year incumbent had represented the Country Party for most of his career prior to losing preselection (at which point he continued holding the seat as an Independent).  On his death the seat was won by his daughter for the third-term Liberal government.


We can see above that the win of Fisher by a fourth-term state government, if confirmed, is an extremely unusual event.  Not only has nothing like this happened in any state for 41 years (which is rather incredible given the depths of unpopularity plumbed by several federal regimes in that time), but similarly strange cases often involved governments replacing friendly independents.  

Discarding the two World War 2 cases and the two where a long-term high-profile incumbent threw in the towel to do something else, we're left with just five cases as exceptional as this or more so.   Liverpool Plains is ancient and information on it sparse, and Kedron might be considered as something like a friendly-fire retirement.  That leaves Bathurst (very long incumbency), Lismore (good luck meets epic tactical error) and Maryborough (exceptional circumstances).  The Fisher win, if confirmed, will be one of the most remarkable state by-election wins by an incumbent government in Australian history.  

It's not enough to just invoke federal drag as a cause of this result, because if it alone explained it then there would have been many other wins like it.  The Abbott government is unpopular, but it isn't extremely unpopular. The David Johnston submarine gaffe, the campaign tactics of the Liberals in heavily targeting Woodyatt, the qualities of the ALP candidate and campaign, the ongoing inability of the SA Liberals to take a trick - all these are likely to be submitted as reasons (or as Mumble would say, Reasons) to explain the result in retrospect.  

A possibly major factor is the desire for stability - voters generally do not much like hung parliaments, and a request from the Liberals to give them a seat so they could try to whip up some chaos with the crossbench wouldn't have been too well received.  (See Carey Moore's list which goes into many of these in more detail.)  Because of the number of factors involved, we shouldn't expect this kind of result to be repeated.

All up though, this (if confirmed) is a hideous result for the Liberals at both state and federal level.  For the seat to have just gone to another independent would have been a minor disappointment, but for it to go to the ALP is a disaster.  Even if Labor somehow manages not to claim the seat, the size of the swing is reason for introspection enough.  



  1. Labor has an advantage over Woodyatt, as Labor lodged a preference ticket under South Australia's weird preference ticket system, while as far as I know Woodyatt has not. This means that a 1 for Labor is formal, while a 1 for Woodyatt is not.

    1. We know that Woodyatt lodged an open HTV card but that is not necessarily the same as the preference allocation submitted to ECSA - someone could lodge an open HTV card publicly but submit an ordered ticket for savings provisions purposes. Is it known Woodyatt didn't do the latter?

    2. I don't quite know. I think the booth displayed card is the ticket, and that was open. The PDF is on the ABC website covering the by-election.