Ballot Order for Blokes

Amy King and I have just put out a paper estimating the effect of ballot order in Australian elections. We look at all federal elections since ballot order was randomised in 1984, which gives us over 500 races (ie. electorate level contests), and over 7000 candidates.

The big surprise for us was to discover that the effect of ballot order is not uniform. If a bloke tops the ballot paper, he gets another 1.4% of the vote. If a woman tops the ballot, she gets no benefit. And 1.4% isn’t chickenfeed – one in ten races are decided by smaller margins than that.

Just for fun, we put together a list of the major party men in elections since 1996 who have had the #1 spot on the ballot paper, and won by less than 1.4%.

  • Kim Beazley, the Labor Party candidate for Brand in 1996 (50.2%)
  • Ross Cameron, the Liberal Party candidate for Parramatta in 1998 (51.1%)
  • Michael Lee, the Labor Party candidate for Dobell in 1996 (50.1%)
  • Gary Nairn, the Liberal Party candidate for Eden-Monaro in 1998 (50.2%)
  • Paul Neville, the National Party candidate for Hinkler in 1998 (50.3%)

Our best estimate is that without ballot order effects, these candidates would have lost.

The paper also comes with a policy recommendation. Randomised ballot ordering is fair before the ballot draw. However, after the draw, it’s unfair that someone might win not because of the quality of their candidacy, but merely because of luck. A better system would be to rotate ballot papers (known to the wonks as Robson Rotation), so the donkey voting effect is evenly distributed. 

The abstract to the paper is below. I’d be happy to discuss the results in comments, though in some cases, you might find your question is answered by a quick flick through the paper (hyperlinked from the title).

Ballot Order Effects Under Compulsory Voting
Amy King & Andrew Leigh
Australia was one of the first countries to introduce random ballot ordering, and is one of the few democracies to make voting compulsory. Using all federal election results since the introduction of random ballot ordering in 1984, we estimate the magnitude and correlates of the ballot position effect in Australia. We find that being placed first on the ballot increases a candidate’s voteshare by about 1 percentage point, or 28 percent of the voteshare received by the typical candidate. Strikingly, however, we find that the ballot order effect exists only for male candidates, who gain 1.4 percentage points when they head the ballot. By contrast, women gain no advantage from topping the ballot paper. Our estimated effects are politically important, since 1 in 10 races are decided by margins of less than 1.4 percentage points. We find no evidence that the ballot order effect has risen or fallen over time. In percentage point terms, the ballot order effect appears to be similar for major party, minor party and independent candidates. Across electorates, the ballot order effect is larger where fluency in English is lower.

Update: The Electoral Reform Society of South Australia have apparently been pushing for some time for Robson Rotation in that state.

This entry was posted in Australian Politics, Uncategorized. Bookmark the permalink.

26 Responses to Ballot Order for Blokes

  1. cam says:

    I did some graphs to see if the Robson Rotation affected churn rate. Dont know if am enamoured with my method, or if the result is concrete.

    The Robson Rotation is an excellent electoral technology, consistent with Australian innovation in that area. I think it should be extended to all state and federal elections.

  2. Cheryl Bookallil says:

    The answer is simple….put the women on top!

  3. Sinclair Davidson says:

    Nice paper. Some thoughts on improvement.
    How about showing the winning voteshare when first compared to winning voteshare when not first? Just for comparison.
    Also why not show a regression controlling for the incumbant? You make the point that Beazley might have lost Brand in 1996, but he was also the incumbant.
    Moving from primary vote gain to two-party preferred comparison makes me uncomfortable. Sure having a higher primary vote is likely to lead to victory, but then why worry about 2PP? Is there any research on winning the primary vote but losing the 2PP? there is a flip side to the ‘big men who won’ data, the ‘big men who have lost’ and ‘ the big women who could have won’ and so on. how many seats won by less than 1.4% where the winner wasn’t first?
    Any view on whether this has, or could, impact on forming government? If it has no impact on the formation of government, is there any pressing need for reform?

  4. Sylvia Else says:

    The implication of the fact that women get no benefit from being at the top of the paper is that even under a Robson Rotation, men get an unfair advantage over women because men will be at the top of some ballot papers, from which they get a benefit, whereas women, who will also be at the top of some ballot papers, do not.

    Perhaps the candidates’ first name’ should be omitted, with just their initials shown, so that the donkey voter gets no clue as to the sex of the candidate.


  5. derrida derider says:

    Our best estimate is that without ballot order effects, these candidates would have lost.

    Err, no. You don’t know the error structure so you can’t predict probabilities that a particular case was higher or lower than the point estimate. And of course even if you could predict the probability in an individual case it is most unlikely all of these would have been euqal to or greater than the point estimate .

  6. Geoff R says:

    isn’t there rotation in Tasmania and the ACT? It would mean that how to vote cards would be impossible, would this impact on levels of formal voting? The women factor is very strange, why would alienated donkey voters vote against women?

  7. Sinclair Davidson says:

    ‘why would alienated donkey voters vote against women?’

    They wouldn’t. Andrew’s result shows lower English levels associated with not voting for women. This could be due to individuals from non-English speaking background being more ‘traditional’ or ‘conservative’ and prefering males to females. Unfortunately, that hypothesis can’t be tested with Andrew’s data (or any other database I know of). Just having initials doesn’t solve the issue – all polling places have photo’s of the candidates.

    The question reminds me that I couldn’t undertand why higher income individuals should be disenchanted with the political process. (Lack of tax-reform perhaps? 🙂 ).

  8. Robert says:

    Geoff’s question is a good one. Why is there a gender difference? Is there any research on the identity of donkey voters — for instance, are they more likely to be male?

    If the Robson system was introduced, wouldn’t it mean that the men still get an advantage (albeit a much smaller advantage)? Surely the fair system would involve ordering the ballot paper so that there was no advantage — which means, in this case, putting a woman on top. The fact that there has been no change in the ballot order effect over time supports this idea. However, it nonetheless offends our sense of fairness.

  9. Robert says:

    Oh, and another thought. You mentioned in your paper that there were some names that you had to look up public records to distinguish their gender. One assumes that the donkey voter is not quite so motivated. How did these “gender neutral” candidates fare?

  10. Pingback: » Blog Archive » Week #26

  11. Anna says:

    Interesting stuff.

    Like Robert, I wonder what happened to the candidates whose first names didn’t make their gender clear.

    I’m not sure that the effect is due to NESB voters avoiding female candidates, though. If your English isn’t good enough for you to be able find out at least the basics about the election, it’s probably not good enough to for you to be able to decide whether a name from the wide range of first names used in Australia signifies a male or a female candidate.

    Perhaps this effect is caused by sexist, English-speaking donkey voters?

  12. Anna says:

    On second thoughts …

    If voters pay attention to the posters outside the polling booth or the pictures on the how-to-vote cards they will be able to figure out the gender of the candidates.

    I’d be interested in ACT results, since posters and HTVs near polling booths aren’t permitted in ACT elections.

  13. Anna says:

    Interesting should be interested above! (Is there a way for me to edit my posts here?)

  14. Andrew Leigh says:

    Lots of great comments, thanks. Some random responses, in no particular order.

    Sinc, we don’t need to control for incumbency (by construction, ballot order is uncorrelated with it), but we could look for differential effects of ballot order on incumbents and non-incumbents.

    Cam, I’m interested in why RR should affect incumbency. The only story I can think of is that incumbents are more likely to hand out HTVs, and RR makes them harder to follow. But I’m not sure even I find this convincing.

    Derrida, the extrapolation seems fine here. We observe what happens to a population when X% receive a random shock, and because it’s random, we can estimate the counterfactual. Of course, there may be heterogeneous treatment effects, but I we deal with the main ones.

    We don’t have enough gender-neutral candidates in this sample, but we’re doing another paper using all elections since federation, and that gives us enough data to look at it (we have nearly 20,000 candidates).

    Anna, I like the summary of “sexist, English-speaking donkey voters”. In return, I’ve fixed your typo.

  15. Geoff R says:

    Beazley was not the incumbent for Brand in 1996 he moved across from Swan.

  16. Sinclair Davidson says:

    Are you sure about Beazley? I thought he moved later – anyway I’m happy to be corrected. On the incumbancy thing, I don’t mean control in a statistical sense, but control in a decision making sense. Some voters may always vote for the incumbant, or against the incumbant (another form of donkey vote). It’s just a thought, the paper makes a contribution as it is (IMHO).

  17. Graeme says:

    Interesting results – especially the gender bias. Clearly it is wrong to say that topping the ballot is always worth 1.4% (even on average). There are numerous instances of micro party and independents, especially in crowded races, drawing first place but polling under 1%. Ballot labelling as well as gender must lead some voters who otherwise scan for the first ‘acceptable’ candidate, to become ‘modified donkeys’.

  18. cam says:

    Andrew, I’m interested in why RR should affect incumbency.

    Dont know. Tasmania has some other electoral technologies which may affect incumbent rates too. IIRC they ban political barracking around the ballot boxes for a couple of hundred metres, which makes handing out ‘how to vote’ literature harder.

    They also dont have bi-elections. If a member leaves parliament, then the candidate with the next highest number of votes gets in. Between that and multi-member districts it means parties are fielding multiple candidates in each seat.

    Maybe those that vote for a party, who dont have HTV literature, are spreading their votes amongst the party candidates more uniformly due to the Robson Rotation.

    I reckon the Tasmanian electoral system is fascinating.

  19. The Speaker says:


    I would be interested to see the result based on party, eg does the ALP benefit from the Donkey Vote than the Liberals ?

    I’ve often wondered what the effect is of the ALP having the longest name on the ballot paper. Does it draw more attention and dominate the ballot paper ? Would this reduce the donkey effect for a Liberal who was at the top ?

    I would also like to see the gender also divided by party. Maybe women on the left of politics do receive some level of donkey support.. ?

  20. derrida derider says:

    Andrew, are you really saying that all candidates at the top of the ballot got exactly 1.4 percent more votes than they would otherwise have got? Surely this is a random effect – in which case you can’t say that all of those who won by less than 1.4% would have lost if they had been at the top of the ballot, because it is probable that for some the effect was less than the average. Conversely, there will be cases where the winning margin was more than 1.4% that were still decided by the donkey vote. /pedantry

    IIRC the old CPA often chose one of the Aarons’ (one of Oz’s great old leftist dynasties) as a candidate, back before the order was randomized.

    The Speaker, if who is at the top is randomly chosen (as it has been since 1984) then there can be no systematic benefit to either party or to a particular gender.

  21. Norm says:

    Beazley moved from Swan to Brand in 1996, as Swan was becoming too marginal for a prospective leader. He won Swan in 1993 with 50.22% 2PP, won Brand in 1996 with 50.23% 2PP. Much safer!

    Cam/Andrew I expect the impacts of RR on incumbency would be different depending on whether its a single or multi member electorate. A single local member would attract more recognition, allowing voters to more readily identify them in a ballot.

  22. Peter Allan says:

    The 1.4% seems to be about right, probably more in non-party local government elections. Robson Rotation should be introduced and the best thing about that is doing away with those damn how-to-vote cards. It won’t happen! The duopoly couldn’t handle it

  23. Graeme says:

    I know it’s ‘just for fun’ (or a press release!) but isn’t there another reason to prevent us saying any individual like Beazley ‘won’ on the donkey vote? The study only predicts ave. first preference benefits. All marginal seats are decided by second, third etc preferences, about which it is very hard to say anything without studying the actual ballots. The person who tops the ballot paper may actually suffer if they receive no preferences from those who consciously vote for a candidate lower on the ballot, then just number their ballot in some ‘donkey-ish’ order.

  24. Sacha Blumen says:

    Andrew, I hadn’t read the full post, and was going to suggest that you’d need something like a completely randomisation rather than just an initial randomisation and then cycling of candidates, but I see you can written that.

    It’d no doubt difficult to look for, but is there a way one can see how many people voted a “pure donkey” ie 1, 2, 3, … down the ballot paper?

    Just thinking that how to votes would be quite different with RR.

  25. Sinclair Davidson says:

    You got a letter from Gary Nairn and a cartoon on the AFR letters page. Fantastic!

  26. Tom Round says:

    The problem with initials-only on the ballot-paper (and I have voted in some University board elections that did just that, presumably for gender-neutral reasons) is that you will almost certainly know the incumbents’ gender – that “J Howard” is John not Janette, and “B Hawke” is Bob not Blanche. Keeping the candidate’s gender secret after they were elected would be impracticable (and conflict with the goal of transparency). Even initials-only ballot-papers work against those voters who actually want to vote for female candidates first (and I’ve known a few women who did that, even when it meant a left-wing feminist voting for a right-wing Young Nationalette because at least she was A Woman).

    Alternative explanation for A1 and A2’s observed 1.4% difference: that there are usually more male candidates than female, especially among the larger, seat-winnable parties. Female candidates still tend to stand out somewhat. (Which cuts both ways… they have to put up with reporters writing “The President looked resplendent in red dress and matching shoes as she announced the nuclear strike”, etc, etc). So – I wonder if female candidates have picked up all the undecided voters they are likely to pick up, just by being female? Whereas because male candidates are still, numerically, the “norm”, a tie-breaker like “first on the ballot-paper” may be a factor among them?

Comments are closed.