Saturday, August 28, 2021

A Record Begging To Be Broken: Labor's Low Winning 1990 Primary

One of the most tedious aspects of Australian polling commentary from time to time is obsessive focus on Labor's primary vote.  Somehow the Labor primary vote gets more attention than the Coalition's when the Coalition is more dependent on its primary vote than Labor is.  In current aggregated polling, Labor has a primary vote in the high 30s (BludgerTrack has it at 37.7%) but there will always be people who say this isn't high enough and that the party needs very near 40% or it can't win a majority. 

There may well be truth in this in terms of projecting from the present day to an election (because oppositions tend to fade and therefore have usually needed big leads at some stage of the term to win) but in that case the primary vote is not telling us anything we couldn't have determined from 2PP polling and its history alone.  The idea that Labor needs a primary vote with a 4 in front of it to win a majority on election day, premised on it only once having won a majority with slightly less, is incorrect.  The fact is we haven't seen a lot about what primary votes Labor wins a majority with these days, because it has only once done so since the rise of the Greens.  This article shows that at recent elections, Labor has missed four in-theory chances to win a majority with a lower primary vote than its record low 1990 majority-winning score of 39.4%.  While total minor party votes remain relatively high, it is very likely that if Labor ever wins federal elections again, it will someday break that record.  The primary-vote-to-win history - based overwhelmingly on elections either run in different minor party environments, or else lost heavily - is meaningless. What matters is whether Labor can assemble a decent enough 2PP result and a good enough vote distribution to win a majority of seats.  

The 1990 Election

The 1990 federal election was unusual, and a precursor for elections to come much later.  Prior to 1990 the minor party vote had usually been below 10%, or where it was higher a large share had often been the pro-Coalition DLP or (during World War II) independents.  In 1977 there was a high Democrat vote, but Democrat preferences split fairly evenly and Labor lost heavily anyway.  All these things meant that prior to 1990 Labor had generally needed close to half the primary vote to win a majority.

In 1990 the minor party vote shot up to 17.5%.  This minor party vote consisted mostly of Democrat and Greens voters, and the rate of flow of Democrat preferences to Labor increased (widely credited to courting of minor party voters over environmental issues, especially by Graham Richardson).  Even so, Labor's poor primary vote saw them narrowly lose the two-party preferred vote (49.9-50.1).  But incumbency often helps with managing the impacts of 2PP losses, and that was enough for the incumbent government to survive with a majority of eight.  Labor had retained a majority despite a primary vote swing of nearly 6.5% against it.

1990 was a forerunner of modern elections where Labor's 2PP is much better than the primary vote gap suggests because of favourable preferences.  However more recently this has happened because very strong 2PP flows from the Greens (high 70s percentages to, at the last few elections, low 80s) have more than cancelled out the flows from right-wing minor parties.  In 1990 there was very little right-wing minor vote.  The Democrat flow to Labor of 63.5%, while much lower than the current Greens flow, was substantially higher than at other elections where the Democrats polled anything to speak of.  (Often it was not much more than 50-50.)

State Level

Labor's 1990 feat was unprecedented at state or federal level.  Prior to 1990 Labor had rarely won majorities with its primary vote below 45% anywhere, achieving some majorities in the 42-44% range through circumstances such as three-cornered conservative contests, ALP breakaways or, in the case of SA 1915, a weird multi-member electoral system.  The pre-1990 records for the lowest majority-winning ALP primary have since tumbled, in some states repeatedly, in every state since 1990:

The pre-1990 records exclude some early Western Australian elections where there were large numbers of uncontested Labor seats.  

Winning a majority off a given primary vote in Tasmania is more difficult because of the state's use of the Hare-Clark system.  What is noticeable is that in every mainland state, Labor has now won a majority with a primary lower than its 1990 winning federal primary.  Each state has its own quirks that have contributed to this, but this should show that there is nothing special about Labor winning elections with primary votes in the high, but not very high, 30s.  

Targets For Majority

I have looked at all the federal elections from 1990 onwards, to ask this question: What sort of primary vote might Labor have won a majority on, given the resulting post-election pendulum?  For these purposes I assume a uniform swing to Labor or against Labor from the actual election results, and treat the swing as going directly from one major party's primary vote to the other, to see what Labor's 2PP and primary targets might have been through the rear-vision mirror.  Of course, this is only approximate (if Labor had done better nationally in any given election then the benefits of that improvement may well not have been distributed evenly nationwide, so the real targets might have been higher or lower) but I think it gives a useful rough idea.  

The chart below contains the following columns:

* ALP PV: The Labor primary vote.

* ALP 2PP: Labor's two-party preferred

* L-NP PV: The Coalition primary vote

* ALP 2S: Labor's share of all votes that went to a major party.  

* Target PV: The primary vote at which Labor would have won a majority assuming uniform primary vote swing from the actual result (and leaving non-classic contests unaffected)

* Target 2PP: The two party vote at which (ditto to above)

* Target 2S: The share of major party vote at which (ditto to above)

Labor majorities are in green with one Labor minority in yellow.  These models assume uniform swing from the final pendulum; Armarium Interrata have done probability based models of the same thing on a 2PP basis (which are on average within 0.45% of the uniform figure, with 1993 and 2019 the biggest differences.  The 1993 difference is because uniform swing leaves Labor with a lot of very narrow losses.)

What does this all say?

1. While Labor won a majority in 1990 despite losing the 2PP vote, this has been an unlikely outcome since.  Reasons for this include the Coalition's advantage on personal votes and marginal seat strategy in 1998, an increase in crossbencher numbers over time and also Labor's vote being inefficiently distributed in 2004 and 2019.  

2. At each election from 1993-2007, a repeat of the 1990 primary vote would probably not have been enough for Labor to win a majority.  Causes of this included a reduced minor primary vote in 1993 and 1996 as the Democrats collapsed, weakened preference flows to Labor in 1998 because of the rise of One Nation, and a more difficult environment for converting 2PP success to seat wins in 2001, 2004 and 2007.

3. However at all of the elections from 2010 to 2019, a repeat of the 1990 primary vote with the swing coming directly from the Coalition would more likely than not in 2010, and almost certainly would in years since, have resulted in a Labor majority.  

4. That's not to say it was as easy on paper for Labor to get a 39.4% primary in all of those years.  In 2010 and 2019 Labor would have needed a substantially better share of the major party vote than in 1990 to get such a high primary.  

5. Indeed in 2019 the deck was heavily stacked against a Labor majority - mainly by Labor wasting swings in inner-city safe Liberal seats while the Coalition did well in its own marginals.  (This is sometimes attributed to demographics, eg education levels, but I estimate that's only half the story.)  One of the reasons the 2019 election was misread is that the actual pendulum at the election was much less favourable for Labor than the pre-election pendulum model suggested.  This is unusual (again see the same Armarium Interrata post that points out how unusual).  

For the record to have been broken in either 2013 or 2019 Labor would have needed to do very much better than it did.  In 2010, the window to break the record was very narrow.  2016 was clearly the biggest missed opportunity to break it, since even a primary vote of 36.2% could have been enough but Labor could only manage 34.7%.

Why has it lasted?

Overall, the reasons the 1990 record has survived so long are:

1. Most elections since 1990 have not been favourable for breaking the record, because of low minor party votes, weak preference flows, or pendulum advantages for the Coalition.

2. The last three elections have been favourable for breaking the record, but Labor's primary vote has been much too low for any chance to do so.

Favourable conditions for breaking the record are minor party votes higher than 20% and preference flows to Labor of above 60%.  If these conditions continue - and there's currently no evidence that they won't - then someday, eventually, Labor will win a majority with a primary vote in the 37-39% range, provided it can even get that high again.  After all, Labor had never won with a primary vote below 45, until they did.  

Even in 2019, where the deck proved so friendly to the Coalition, had Labor repeated its 1990 primary vote with the difference coming off the Coalition, Labor would have won the 2PP 54.6-45.4.  The swing would have taken out Peter Dutton, Michael Sukkar, Greg Hunt and very nearly Josh Frydenberg as Labor won the election with a seat result of something like 88-56-7.  

Given this it is completely absurd to maintain that Labor needs to match its 1990 primary to win a majority.  


  1. For the best example of winning from a low primary vote with lots of preferences:

    Under Bolte the Liberal Party* in Victoria won majorities 6 election in a row with primary votes bellow 40% each time, as opposed to the general trend of governments not being repeatadely elected with low primary votes. In 1961 the LCP only got a 36.4% primary but won 39 out of 66 seats. In 1970 the Liberal Party got a 36.7% primary but won 42 out of 73 seats, with the Country Party directing preferences to the ALP (but the ALP not reciprocating in some Country Party seats, sending them to the Liberals).

    *Called the Liberal Country Party 1948-1965.

  2. One more thing to note - the primary figures in this post are very much an "upper bound" on the minimum Labor primary needed for a majority.

    Let's say, for example, the 2016 election was held a few months earlier and Australia was hit by severe bushfires, causing interest in climate action to surge and the Green vote to rise to 13% (similar to the figures Newspoll found in Dec 2019).

    If the rise in the Greens came primarily at the expense of the Coalition (with little change in the Labor primary) and the GRN -> L/NC preference flow stayed stable, then in our scenario Labor would have won 51.9% of the 2pp and almost certainly a majority with just 34.7% of the primary vote. In fact, they would still have room to lose 0.8% of their primary to the Coalition - and still win a majority anyway!

    All this goes to show how ridiculous the obsession over primaries is. Unless a minor party is winning high enough primaries to win actual seats (very difficult - just ask the Green or check in with SA-BEST), 2pp is, for now, still the key determinant of electoral success.

  3. Oh, and one more thing you might want to deal with in this piece:

    Claim by the same person (apparently an Assoc Professor?) that the Greens' presence in current-day elections is equivalent to the Democrats' historical presence.

    1. Ta; I hinted at it but have expanded the relevant section slightly. Wallace's tweets are so riddled with errors that if I addressed them all I could probably write an article doing so every week!