How Much Does A Home State Federal Leader Matter In That State?

This piece came about, indirectly, because the Courier-Mail was printing silly sentences again.  A piece commenced with the following lines:

"Kevin Rudd has refused to reveal whether under-pressure Opposition Leader Anthony Albanese can win Queensland despite the former Prime Minister joining him for day one of a whirlwind ‘jobs tour’ across the state. 

[..] 

Mr Albanese was joined by Mr Rudd on the pre-campaign hustings at Southbank – however the former Prime Minister declined to comment when asked by The Sunday-Mail about how or whether Labor could win the state."

Rudd claims he declined to comment because it was the Courier-Mail rather than because of the question.  Whatever the facts in that regard, two things stood out to me in the opening line. Firstly there's the idea that the former Prime Minister is a divine revelator who can exclusively "reveal" facts  about the prospects of Opposition Leaders in his state.  Secondly, the idea that winning Queensland is a realistic or relevant goal for an Albanese-led ALP.  After all, Labor currently needs an enormous 8.44% swing to win the 2PP in Queensland, and has only won the Queensland 2PP three times since the Second World War - 1961, 1990 and 2007, all of them by a whisker

What Labor should be aiming for in Queensland next time around is to lose the state by much less (if it can), and hence win back seats.  Indeed if by some small miracle it could lose the 2PP only 48-52 instead of 42-58, it might recover about five seats, which all else being equal would be enough for minority government.  I am not suggesting this will happen.

I did suggest that winning Queensland might be worth talking about if Labor switched to Jim Chalmers ("Jim who?" to almost every non-tragic south of the Tweed at this stage, I suspect) as leader.  That made me wonder though: just how big an effect in federal elections is it for a party to bring in a leader from a particular state?  

All else being equal, if a leader boosts their party in their home state we would expect the following:

1. The swing to the party in the home state that has gained a party leader should be higher than the national swing to the party.

2.  The swing to the party in the state that is no longer a home state for the leader should be lower than the national swing to the party.

(For swings against, the swing to the party is a negative number).

There is one reason why this might sometimes not be the case - a party could deliberately change leader to a leader from a state where it thinks it is facing an especially adverse swing to try to avoid that swing.  However I don't think that's been a major feature of past leadership changes.  

Here's what the data show ...

Data Table

For elections from 1922 onwards 2PP figures are available in a way that makes it possible to calculate a swing from the previous election's 2PP, both nationally and on a state by state basis.  I've used David Barry's table for this purpose.  

There are 26 cases of the state of a party leader changing between elections, though in the case of the Nationalists in 1922 this involves Billy Hughes switching from a Victorian seat to a NSW one.  The other 25 all involve changes of leadership.  1922 was one of two elections where both sides switched from having a leader in Victoria to having one in NSW, the other being 1993.  1946 also saw both sides have a leader from a new state, but without overlap.

In the table below (conservatives on the blue side, Labor on the red side), the columns are, from left to right for each side:

Nat: the national swing to that side

Old: the state that the party's leader was from at the previous election

Swing: the swing for the party in that state

New: the state that the party's leader becomes from at the current election

Swing: the swing in that state

Old-N: the swing in the old state minus the national swing

New-N: the swing in the new state minus the national swing

New-Old: the swing in the new state minus the swing in the old state.  This figure is underlined to highlight cases of acrimonious leadership transfer (spills and resignations under public threat of spill):

(Click for slightly larger clearer version.  The table has been hand-compiled and I don't guarantee it is entirely error-free, but will revise and edit if any errors are detected.)

Comments on the results

The following are my comments on the above results:

1. The effect for a state gaining a home-state leader is quite weak.  The Coalition has outperformed its national swing by around 1% in the states that became the state of its leader; Labor has done so on average by only 0.4%, though this rises to 1% for elections after 1961. All Labor leaders since then except for Bill Shorten have outdone the national swing in their home state at the first election since the switch.  If the two elections with clashing leader-state changes (1922 and 1993) are dumped from the sample, Labor's average goes up to 0.6%.  (I'm not even sure the effect is statistically significant.  If it is I expect it's only weakly so.)

2. The reverse effect for a state losing a home-state leader may or may not be a thing.  Across the sample as a whole, it either isn't a thing or is tiny, but for elections after 1950 it is worth about 1%.  It's possible that messy 2PP swing patterns (including party splits) in elections from the first half of the 20th century have obscured it before that.  See also point 5 below.

3. The leader's state never makes an enormous difference in that state.  There is no case of the state of the new leader recording a swing of four points better than the national swing.  A boost of a few percent in any state isn't to be sneezed at as in cases it will be worth several seats, but it still isn't 10%.  

4. There's some evidence the new-state effect could be stronger for leaders outside NSW and Victoria.  Too small a sample size to be confident, but it's what would be expected because of the greater novelty value for residents of those states.  The seven cases involving a change to a leader from Qld, WA or Tasmania yield the Coalition's second and third highest New-N scores and Labor's second, third and fourth.  The average New-N score for these seven changes is 2.15%, nearly three times the overall average.

5. Leadership stoushes between rivals from different states may turbo-charge state effects.  Again, small sample size and very hard to be sure these aren't caused by other things.  Very high New-Old scores are seen for Labor in 2010 (Julia Gillard deposing Kevin Rudd, Labor doing very well in Victoria and very badly in Queensland), and the Coalition in 1990 (Andrew Peacock deposing John Howard).  2013 was a reverse of 2010 and showed the same effect in reverse but to a smaller degree.  1972 (McMahon-Gorton) and 1983 (Hawke-Hayden) both also have above average New-Old scores, but 1993 (Keating-Hawke) lives up to its record of breaking almost every rule (probably because of the massive influence of the Victorian Kennett government on that election outcome, rather than the fact that the Coalition also changed the state of its leader).  The 1990 and 2010 cases may both be partly coincidences - Labor was on the nose in Victoria and Queensland respectively at state level in those years, and that may have affected the federal results.  

Overall, the impact of picking a leader from a particular state in that state is not more than a few points in that state, and it could be cancelled out (or worse) if a leader from a bigger state has to be messily thrown under the bus to get there.   

These historical results do not suggest that picking a leader with an eye to boosting the party in a particular state would ever be a good strategy.  It seems it is best to be far more concerned with how a leader will be viewed across the country, and if the answer is well, then their home state will take care of itself.