Home > General > Swinging voters?

Swinging voters?

February 3rd, 2004

In today’s SMH, Gerard Henderson repeats all the standard claims about swinging voters, for example

Elections in Australia are invariably decided by people who in most other democracies would not vote, or who, if they choose to back a minor party or independent, would not express a preference for either of the major parties. This is due to the unique federal electoral system, which comprises compulsory and preferential voting – introduced in 1924 and 1918 respectively.

In other words, it is the essentially uncommitted and/or uninterested – living in marginal seats – who decide election outcomes in Australia … Since the end of the Pacific War, the Government has changed hands only five times – 1949 (Menzies), 1972 (Gough Whitlam), 1975 (Fraser), 1983 (Hawke) and 1996 (Howard). The outcome in each case was determined by the change in allegiance of essentially non-political voters, along with a proportion of newly enrolled electors. 

There’s a trivial sense in which all of this true, just as the outcome of a cricket match [at least if it isn't drawn] is always decided by the last ball. And in a constituency-based system with only two parties capable of winning seats, it’s broadly speaking correct that only marginal seats count – however, swings are more variable these days so more seats are marginal.

Apart from that, however, Henderson’s analysis is (to the extent that it was ever valid), totally out of date. The implied picture is one in which most voters, and all who take a serious interest in politics, are committed supporters of one major party or the other, in roughly equal numbers. Hence, the remainder, the ‘swingers’ determine the outcome.

This was always a problematic viewpoint. As Henderson’s own family experience during the Split (he has an interesting piece on this in the latest Sydney Institute newsletter) people can and do make permanent changes in their political allegiance. A conversion of this kind is far more valuable to the gaining party than winning a swinging voter for a single election.

More importantly, the number of “rusted-on” major party voters has declined drastically. Both sides have recorded votes of 35 per cent or below in their worst recent outings, which puts an absolute upper bound of 70 per cent voters committed to one party or the other. But even within this group, there are almost certainly some who changed over time, or voted on a specific issues. For example, Keating lost lots of hardcore Labor voters in the 1996 election, but he undoubtedly gained some “cafe latte” votes with his cultural agenda – many of these will not be so attracted to Latham, while the former hardcore may return.

Henderson is particularly unsatisfactory when it comes to the minor parties. He notes that most Green preferences go to Labor, but there’s a big difference between getting a primary vote and getting 70 per cent of second preferences. To be precise, on these numbers the average Green vote is worth 0.4 first preferences votes to Labor.

More importantly, since minor party preferences are now crucial, what basis does Henderson have for equating minor party voters with “the uncommitted or uninterested”? The decision to vote for a minor party means that the voter has rejected both of the obvious choices, which normally implies some degree of interest in the process. It’s hard to argue that Democrat and Green voters as a group are less interested in politics than major party voters, whatever you might think of the results of their interest. Even One Nation voters, who are on average closer to Henderson’s idea of the apolitical swinger, were expressing a (negative) interest in political ‘business as usual’ when they abandoned their traditional allegiances to vote for Pauline.

Categories: General Tags:
  1. cs
    February 3rd, 2004 at 16:06 | #1

    Good work mate. Very enjoyable … should be a Fin column there.

  2. February 3rd, 2004 at 16:15 | #2

    Yeah,

    I reckon its pretty good analysis as well.

    Don’t know why Radio National keeps using Henderson.

    They need a resident punk conservative commentator in pinstrips and ear rings.

  3. JimB
    February 3rd, 2004 at 16:36 | #3

    Is this Henderson’s funny roundabout way of implying that Howard is washed up?

    As I recall Howard’s last reelection was deemed a substantial democratic triumph whereas his apparent upcoming defeat is merely a sad reflection on the fickle average aussie idiot voter.

  4. February 3rd, 2004 at 17:59 | #4

    John,

    I don’t understand the calculations behind your statement: “there’s a big difference between getting a primary vote and getting 70 per cent of second preferences. To be precise, on these numbers the average Green vote is worth 0.4 first preferences votes to Labor.”

    While getting raw first preference votes may count for some subsidiary things (like AEC funding?), how is a second preference vote more dilute than a primary vote when it comes to the main game?

  5. John
    February 3rd, 2004 at 18:21 | #5

    Suppose 100 people vote Green, 70 giving Labor their second preference, and 30 Liberal. When the preferences are distributed, Labor has a net gain of 40.

    Dividing through by 100 gives the result that, on Henderson’s assumptions 1 Green vote is worth 0.4 first preference votes.

  6. February 4th, 2004 at 14:44 | #6

    John, I think your numbers are a pea and thimble trick.

    But using your equation, the Libs also suffer a net .3 of a vote loss, which net-net means .7 vote to Labor. As intuition would suggest.

    But I don’t think much of Henderson’s analysis either. It’s true that the outer suburbs tend to be marginal and change from side to side (though they’re not as marginal as the regions), but no real reason to believe people are less engaged there than anywhere else.

  7. John
    February 4th, 2004 at 16:06 | #7

    Mumble, I don’t follow your argument, but I should note that I didn’t specify the problem as rigorously as I might have.

    Let’s begin by taking 100 new voters. If they all vote Labor, that’s a gain of Labor to 100. If they all vote Liberal, that’s a gain to the Liberals of 100. IF they divide their votes evenly, that’s zero. If they vote Green and allocate their preferences as above, that’s a net gain to Labor of 40.

    So each new Green voter is with 0.4 votes to Labor, on a scale where 0 means that the voter is equally likely to go to Labor or Liberal and 1 means the voter gives Labor first preference every time. [On this scale a Liberal vote counts as -1 for Labor].

    The second case, referred to by Henderson is where voters switch from Labor to Green. The same analysis as above implies that for each voter who does this (and assuming 70 per cent allocate Labor second preference) Labor loses 0.6 votes, expressed in terms of the final margin between the parties.

    If I interpret you correctly, you’re saying that things are even worse from the Liberal viewpoint if a potential or actual Liberal voter goes to the Greens and this is quite true. The analysis above implies a new Green voter is worth -0.4 to the Liberals, and a switch from Liberal to Green costs them 1.4 votes (I’m expressing this in majority terms so a switch from Liberal to Labor costs the Liberals 2 votes).

    It’s easy to check that everything cancels out if 30 per cent of Green voters are former Liberals (or if 30 per cent of new Green voters would otherwise have been Liberals). This would salvage Henderson’s argument, but he explicitly focuses on Green votes taken from Labor.

    A further possibility is that the second preferences are highly correlated with the former party. My very casual observation suggests that this is not the case and that, having switched on first preference votes, people are often willing to do the same on second preferences, if not immediately then after a couple of elections.

  8. Andrew Norton
    February 4th, 2004 at 16:50 | #8

    I checked in the 2001 Australian Election Survey, which asks about party ID and about interest in politics. Taking “no interest” in politics to be people who say they have not much interest or no interest the rough results are Lib – 18%, Lab – 23%, Nat – 23%, Dem – 21% Green – 25%, One Nation – 41%, No party ID – 37%. The numbers in the sample of the minor parties are a bit small to conclude too much from all this, but most minor parties have no interest voters in the normal range.

  9. February 4th, 2004 at 19:13 | #9

    Dear John,

    There’s something fishy about your equation, to be sure to be sure, and I propose this hypothetical:

    There’s a federal election in November 2004, at which a field of candidates runs for the Queensland seat of Pedantry.

    After full distribution of preferences, the candidate called John defeats the candidate called Peter by to 70,000 to 30,000.

    “Congratulations”, says a reporter to John. “You got 70% of the two candidate preferred vote.”

    But John is crestfallen. “No”, replies he, “I received a net 40%, well below a majority. I cannot claim victory.”

    John thanks his supporters, apologises for letting them down, and then goes home.

    So debilitating has the experience been he never again puts up his hand for public office.

    End of story.

  10. John
    February 4th, 2004 at 21:01 | #10

    To pick just one of the fatal problems with this example, it’s easy to prove that this result can’t arise on the assumption that minor party preferences are distributed 70 per cent to John and 30 per cent to Peter.

    Check it out – remember of course, that the third party candidate, let’s say Luke, must have less primary votes than either John or Peter. It’s easy to show that this implies that Luke can have polled no more than 23 per cent of the first preference vote, implying that John can have received no more than 16 per cent of the vote as second preferences from Luke. But this means that John must have received at least 54 per cent of the first preference vote, so preferences wouldn’t be distributed.

    Allowing for more than one minor party candidate doesn’t change this at all.

    Even if you suppose preferences are fully distributed just for the heck of it, why would John behave as you say when, as has just been demonstrated, he must have had an outright majority of first preference votes.

    More importantly, this example has nothing to do with the point I’m making.

  11. February 4th, 2004 at 22:48 | #11

    Also more importantly, whether Henderson is conceding or not, the writing is on the wall for JWH. Latham is getting all the atttention he needs and will win an election that even Crean could have won.

  12. February 5th, 2004 at 00:47 | #12

    Yeah wbb … I’m prepared to entertain the possibility that we have another drovers dog thing here … the Hawke-Latham comparisons are rich and as yet unwritten.

  13. February 5th, 2004 at 06:56 | #13

    Can we wait and see what the punters say? I haven’t seen a poll showing Latham doing any better than his predecessor in two party preferred terms.

    Speaking of which! I still accuse John of double counting. Let me put it this way.

    The race is for 50% of the vote plus one. Given any number that is the total formal vote, 50% plus one is also a known number.

    If ten votes are distributed, and I as a candidate get 7 of them, then I’m seven votes closer to that magic number. Ok, my opponent is 3 votes closer, but to say my progress has only been 4 votes would be wrong. I’ve moved forward by seven votes. That’s worth .7 of the situation if I’d gotten all ten of them.

    PS. Latham isn’t Bob Hawke’s bootlace.

  14. February 5th, 2004 at 10:01 | #14

    John,

    I’m not fully behind Mumble’s examples, but I’m still not convinced by your figures, either. (I should caveat at this stage that I am close to a mathematical illiterate – in those share-house scenarios where A owes B and C, C owes B, and D owes A, I used to always insist on the cash being passed around the group, instead of D paying A and C (or whatever) to settle the whole thing).

    I can see your logic, and it’s impeccable down to a 50/50 preference split being a “zero” vote gain for both major parties. But I see problems to do with a 70/30 split being a *negative* for the less popular party.

    Here’s my example: the Libs get 49% of the primary vote; Labor 46%, and the Demogreens 5%. Demogreen preferences are split 70/30 Labor/Lib. Clearly, these *negative* (?) value prefs for the Libs are nonetheless going to get them over the all-important, two-party-preferred, 50.00001% of the vote “line”.

  15. John
    February 5th, 2004 at 20:59 | #15

    OK, one more time.

    In the example given by Paul, the Libs start off with a lead of 3 percentage points. Once the preferences are distributed that lead has fallen to 1 percentage point, which is still enough for victory.

    To venture onto the risky ground of analogy, suppose a team is ahead by 5 goals at three-quarter time and is outscored 7 goals to 3 in the final term, kicking the last goal just before the siren. You could say that the final quarter won them the game, since the winning goal was kicked in that quarter. Similarly, as I noted in the original post, you could say that the last ball in a game of cricket was the decisive one.

    I’d prefer to say, in the circumstances above, that the last quarter was equivalent, in all respects relevant to the outcome to one in which the other team kicked four unanswered goals.(It’s not quite equivalent in terms of the impact on the percentage ranking, but in the political case winning margins don’t matter.)

    At this point, having given the last word, I’m going to (ab)use the power of the blogger and declare this discussion closed.

  16. February 6th, 2004 at 08:22 | #16

    This discussion amply demonstrates the lack of clarity of “fair” systems in general use today. That is, the lack of obviousness of what is going on, that distances the voters (and I have other problems with the systems too).

    In my “humble” opinion, having looked over the crop, the best/least worst single system is cumulative voting, which requires multi-member seats. It has rarely if ever been adopted freely by politicians, boards of directors, or whatever, though it has sometimes been imposed by courts with great effect in improving things. Of course, it has only been applied on a small and/or local scale, so that may not extrapolate to whole systems.

    And, as I’ve mentioned elsewhere, I believe that a proper system should involve two houses working off very different approaches to representation, so as to yield truly independent second opinions, and with no tie-break but rather a way of going to the people to resolve deadlocks (so as to bypass or at least mitigate the problems revealed by Arrow’s Theorem).

  17. February 6th, 2004 at 12:22 | #17

    “equating minor party voters with “the uncommitted or uninterested” is, to borrow a John Howardism, offensive. It’s been remarked on before that there is a strong tendency for Australians to split their vote between the lower house and Senate in a deliberate ploy not to let the successful party wield too much power. They also know that a vote for a minor party is unlikely to put that party into power but is morel likely to make the next government give a little more attention to the minor party’s concerns than they otherwise would. Unthinking? Mate, some of us think too much. Henderson, get stuffed.

  18. February 6th, 2004 at 23:41 | #18

    I beg Pr Q’s indulgence but this numerical example may explain why his maths correctly depict the situation.

    Imagine a preferntial system where there are only two parties, ALP and LNP, in an electorate of 100 voters.

    The vote is evenly split:

    ALP 50%

    LNP 50%

    Since second preferences can only be distributed to the one other competing party, they cancel out. This leaves the parties tied.

    Next imagine that both parties lose 5% each of their voters to a third centrist party, the DGP. These third party voters are slightly left wing in orientation, so they lean slightly in the distribution of their preferences to the ALP: ALP 0.6/LNP 0.4.

    The partisans of the main party’s always second prefer the centrist DGP, so they uniformly distibute all their second prefernces to the DGP. They always distribute their last preference to their respective ideological polar competitors.

    ALP voters

    primary 1st ALP 45%

    distribute 2nd DGP 1.0

    LNP voters:

    primary 1st LNP 45%

    distribute 2nd DGP 1.0

    DGP voters:

    primary 1st DGP 10%

    distribute 2nd ALP 0.6/LNP 0.4

    This gives the ALP 6 votes to LNP 4 votes after preferences are distributed. Thus the new figures are:

    ALP 45 primary + 6 DGP secondary = 51 votes

    LNP 45 primary + 4 DGP secondary = 49 votes

    The result of a 0.6/0.4 distribution of third party preferences to the ALP, from a default position of an even share of the primary vote between the major parties, is to give the ALP a net gain of 2 votes.

    This result accords with Pr Q, not Mumbles, conclusions. Mumbles error appears to be in making the assumption that votes are exhausted in gross plural, not net prefererential, distributions.

Comments are closed.