How much is that vote in the window?

$28,000, if the Coalition’s Roads to Recovery program is anything to go by. In a piece in today’s AFR, I look at the relationship between Roads to Recovery spending over the last election cycle, and the swing towards the Coalition (2004 two-party preferred vote minus 2001 two-party preferred vote).

It turns out that the relationship is positive and significant. For the statistical wonks, here’s the regression across 144 federal
electorates.*

Swing to Coalition = Roads to Recovery spending ($M) + Constant
                                          0.043 (0.018)                          1.53 (0.34)
                                          P=0.021

In other words, $10 million buys a 0.43% swing, which works out to about $28,000 per vote.

Of course, this isn’t conclusive evidence of vote-buying, but it does seem to point that way. If I were confident that this sort of program was good value for money, I might be less concerned, but I found the departmental benefit-cost study pretty unpersuasive. I suppose that if I were a passionate believer in voter rationality, that could also give me reason to be unconcerned, but my read of the evidence is that Australian voters are at best quasi-rational.

Update, 14 March: In a co-authored letter to today’s AFR, Tim Fry and Sinclair Davidson point out that pork-barreling doesn’t account for all of the Coalition’s swing. This is obviously true – but the fact that one roads program doesn’t explain the whole Coalition swing doesn’t mean that it’s not substantively important (would we have expected a single program to explain why the Coalition was re-elected?). By way of analogy, I’m sure Davidson would agree that even although taxes might be a small factor affecting work effort, tax reform can still matter.

* But wait, I hear you cry, there are 150 electorates, not 144. Yes, but in four electorates, I didn’t have road spending data; while two – Bonner and Gorton – didn’t exist in 2001, so we can’t easily calculate the swing.

This entry was posted in Australian issues. Bookmark the permalink.

6 Responses to How much is that vote in the window?

  1. Sinclair Davidson says:

    Hmm, not sure about this. Is the dependant variable a change variable or a total spend? If we think back to the election, very few seats changed hands, so if the coalition won the election, and spent more money in coalition seats anyway, then don’t we expect a positive coefficient anyway?

  2. Sinclair Davidson says:

    Sorry, don’t want to be timed out.

    also wouldn’t we expect more road expenditure in larger constituencies which tend to be held by the coalition as well?

    so have you size adjusted the expenditure?

    any other controls that could be added?

  3. Andrew Leigh says:

    Sinclair, the depvar is indeed a change variable. Swing to coalition = coalition vote in 2004 minus coalition vote in 2001.

    We could control for size if we had a convincing story as to why the Coalition would’ve been expected to do particularly well this time (compared to last time) in big electorates. I’m sure you and I could construct such a story, but it certainly hasn’t been prominent in the post-election analyses.

  4. Sinclair Davidson says:

    Sorry, Andrew – I should read what I type. the INdependant variable – is it a change variable or a total?

  5. Andrew Leigh says:

    Sinclair, it’s basically a change variable. The R2R funding was meant to be a top-up to existing road funding. It is conceivable, however, that some councils cut back their road funding to account for the new R2R money.

  6. Sinclair Davidson says:

    Tim has done something similar with the jobs network data, and finds a similar result. Our point related to your summary statement, (from memory) the coalition paved the road to the lodge. It is a great line but not consistent with your regression. We reckon the $28,000 is a ‘small’ impact.

Comments are closed.