Do state governments drag their federal counterparts down?

Do state governments drag down their federal counterparts? It’s a common theory among Australian political commentators. The 2016 election suggests the theory might be right.

In the 2016 federal election, the four states/territories1 with non-ALP governments2 swung towards the ALP by an average of 5.2 percentage points. In the states where Labor held power, voters only swung towards the federal ALP by 2.6 percentage points, on average. The average swing towards Labor was bigger – 2.6 percentage points bigger – in the non-Labor states.

The 2016 result bolsters the view that state goverments drag down their federal counterparts. The Coalition fared worse in the non-Labor states, and federal Labor fared worse in the states where they held government. What about in the previous election?

2013, too, is consistent with the story. Every state swung against the ALP, but the swing was bigger in the Labor states. The swing was 3 percentage points worse for the ALP in the Labor states, not that far off the 2.6 percentage point difference in the 2016 election.

If we only looked at the past two elections, the story would be reasonably compelling. But the theory seems a lot less persuasive if we look at prior elections. In 2010, the Labor states swung more heavily towards the federal ALP than the non-ALP states, contrary to what the commentators’ theory predicts. In 2007, all states had Labor governments and they all swung towards the federal ALP; in 2004, Labor was in power in all the states and they all (bar the ACT) swung towards the Coalition.  If we go back further in time, the picture gets even murkier.

The last two elections have lent support to the theory that state governments drag down their federal counterparts. The swing to the federal ALP was bigger, on average, in states with a non-ALP government than in ALP states. But if we look back through history, we see plenty of examples where the opposite was true, like the 2010, 2001 and 1998 elections.

On average, if we include at every federal election since 1951, states with Labor governments have swung against the federal ALP by 0.18 points, while non-ALP states have swung towards federal Labor by 0.44 points. That’s about a 0.6 percentage point difference – which is not nothing in a closely poised electoral contest – but it’s clear if you look at the distribution of swings that these averages mask a pretty broad range of outcomes. There have been Labor states that swung towards the federal ALP by nearly 10 percentage points (WA in 1983) and those that swung against Labor by more than that amount (Tasmania 1975); similarly Victoria in 1955 was a Coalition state that swung against the ALP (ie. towards the federal Coalition) by around 10 points.

None of these charts provide a comprehensive answer to the question of whether state parties drag down their federal counterparts. We can see that, on average, the swing towards the ALP is about 0.6 points bigger in non-ALP states. But that doesn’t necessarily mean that the state governments are causing these swings. There are, quite obviously, many other factors at play in determining elections.

To get a better answer to the question of whether states drag down their federal counterparts, and if so how big the effect is, we’d need a model of elections that included a range of other relevant factors.3 If the effect remained after statistically controlling for other things that effect elections – like economic conditions, and the length of time that the government has been in office – then we could have more confidence that there’s some there there. As it is, the theory remains quite plausible, but to me it’s unproven.

EDIT: For anyone who’s interested in looking into this issue further, here’s the data I used. It combines AEC data (on the 2PP vote by state) and UWA Elections Database data (on which party was in power at state level).

 

 

  1. I’ll just call them ‘states’ from here in…
  2. I’m calling them ‘non-ALP’ to cover arrangements like the WA not-quite-coalition under the Barnett Government, and the Newman LNP Government (not a coalition), etc.
  3. Even then, we’d have the old correlation/causation problem beloved of internet commenters everywhere.