One of the many things that worked in Labor’s favor in the 2007 election was the fact it did rather better than its opponents in the draw for ballot paper positions. As I reported at the time, out of 41 at-risk Coalition seats, Labor improved its position in 16, but went the other way only in six.

On Friday the Australian Electoral Commission conducted the ballot draw for this year, and it’s much more even. Out of 48 marginal seats (Labor-held up to 6% and Coalition-held up to 4%), there are 12 where Labor had the benefit of the donkey vote in 2007 but has lost it this year, and 13 that have moved the other way.

Here is the list, with margins (asterisks denote seats that have notionally changed party due to redistribution). Donkey vote shifting to Coalition:

  • Bass (TAS), ALP 1.0%
  • Bonner (QLD), ALP 4.5%
  • Brand (WA), ALP 6.0%
  • Corangamite (VIC), ALP 0.9%
  • Deakin (VIC), ALP 1.4%
  • Dobell (NSW), ALP 3.9%
  • Flynn (QLD), ALP 2.2%
  • Forde (QLD), ALP 3.4%
  • Hasluck (WA), ALP 0.9%
  • Leichhardt (QLD), ALP 4.1%
  • Paterson (NSW), LIB 0.6%
  • Wentworth (NSW), LIB 3.9%

Donkey vote shifting to Labor:

  • Bennelong (NSW), ALP 1.4%
  • Brisbane (QLD), ALP 4.6%
  • Cowan (WA), LIB 1.3%
  • Cowper (NSW), NAT 1.2%
  • Dunkley (VIC), LIB 4.0%
  • Gilmore* (NSW), ALP 0.4%
  • Herbert* (QLD), ALP 0.03%
  • Macquarie (NSW), ALP 0.3%
  • McEwen (VIC), LIB 0.02%
  • Moreton (QLD), ALP 6.0%
  • Robertson (NSW), ALP 0.1%
  • Ryan (QLD), LIB 1.2%
  • Swan* (WA), ALP 0.3%

How important is this? Well in 2007, of the 16 key seats I named, the 11 where Labor gained from the donkey vote swung to it an average of 6.8%, compared to the national average of 5.4%. The five seats where the advantage moved the other way swung by an average of only 4.1%.

That’s not a large sample, but it lends support to the view that the donkey vote is worth something like 0.7% (that is, if it shifts from one side to the other it produces a difference of twice that, about 1.4%, from the average).

The leading recent academic work on the subject is a 2009 paper by Amy King and Andrew Leigh, Are Ballot Order Effects Heterogeneous? Their headline number for the donkey vote is 1%, but what they’re measuring is slightly different from the two-party advantage, and some of their analysis suggests a smaller number might be more accurate.

The donkey vote also varies with things like the number of candidates and the level of education in a seat, but we’re probably not far off the mark if we say that the party gaining the advantage in the seats listed above will do on average a little over 1% better than it otherwise would have. In a close election, that’s well worth having.