Possums Pollytics

Politics, elections and piffle plinking

The Proportional Application

Posted by Possum Comitatus on October 5, 2007

In our last Newspoll breakdown, we divvied up the seat swings according to the Newspoll estimate of three seat types, and by State for 139 seats across the country. That gave some people psephological indigestion because the results didn’t look like an election would at the margins of safe vs marginal seats. Fair enough, but it was never meant to – it was a breakdown of Newspoll to the best resolution available from the data.

This time, we’ll actually set out to create an approximation of what the Newspoll data was saying that isn’t limited by the data resolution, using basic probability theory and an assumption of uniform state swings. We’ll do today what some people thought we were doing the other day.

To start with, the TPP swing is calculated as the percentage change in vote of the two major parties. Hence if a seat moves from 50% to 55%, it’s had a 5% swing.

If we have 3 seats that are, in ALP/Coalition terms: 40/60, 50/50 and 60/40 and the average swing in these seats is 5%, if the people that are changing their vote are drawn randomly from those three seats, we would expect that the seat with a higher proportion of coalition voters will have more Coalition votes that changed, and the seat with less coalition voters would have less coalition votes that changed.

This is simply as a result of a random distribution of Coalition voters changing their votes. As a result, the 60/40 seat would have (0.05*40*2)-40 coalition voters = 36, and (0.05*40*2)+60 ALP voters = 64.

Doing this for the 50/50 seat we end up with a TPP of 55/45 and for the 40/60 seat we end up with a TPP of 46/54. The seats with a larger number of coalition voters swing more, on a given uniform swing to the ALP (here 5%) than seats with a smaller number of Coalition voters. This is played out in nearly every Newspoll marginal seat/safe government seat/safe ALP seat estimate and most elections.

If we apply the various State swings using the same methodology to the 139 seats we are looking at here, we end up with the following

  NSW Vic Qld SA WA
Actual Newspoll Swing 9.2 11 9.1 9.4 4.4
Calculated Swing 9.4 11.2 10.4 10.2 4.9
Ratio 0.9747 0.9826 0.8759 0.9215 0.9029
Adjusted Swing 8.97 10.81 7.97 8.66 3.97

The Calculated Swing is (using NSW as an example) what the average swing in NSW is calculated as after we’ve applied the Newspoll estimate to the individual seats using the methodology outlined above. Notice how it’s higher by a bit across the board? That’s the variation in the uniform swing interfering. So we can adjust the swing downward for this by multiplying the Actual Newspoll Swing by the ratio of the Actual Newspoll Swing to our Calculated Swing. This gives us our Adjusted Swing. When we apply our adjusted swing to all 139 seats, the State averages of the seats all tally up to the actual Newspoll state estimates. By doing this, we are choosing to minimise the size of the swing we use to account for the small variance in the random distribution of changing voters that Newspoll is picking up in their polling.

We can see how the minimising of the swing plays out by comparing our results in three seat types against the Newspoll results for the three seat types:

  Newspoll Swing Our Swing
Marginal Seat Swing 8.3 8.14
Safe Government Seat Swing 11.6 10.18
Safe ALP Seat Swing 7.1 6.83
National 8.8 8.78

We are smaller in every case of the three seat types, but bang on the State and National average swings. So now we have our conservative estimate that is actually well within the Newspoll estimates, we can apply that to the 139 seats in our list to get an approximation of how the election results would have looked had an election been held between July and September. Remember, as we are using a slightly smaller set of swings than Newspoll, this is a slightly conservative estimate. Instead of all 139 seats, we’ll just look at those seats where the ALP TPP vote was estimated to be above 48%

Division State 04 Election Current TPP Estimate Swing
Indi Vic 33.71 48.0 14.3
Cook NSW 36.72 48.1 11.3
Kalgoorlie WA 43.7 48.2 4.5
Forde Qld 38.48 48.3 9.8
Hume NSW 37.16 48.4 11.3
Fisher Qld 39.02 48.7 9.7
Ryan Qld 39.58 49.2 9.6
Leichhardt Qld 39.74 49.3 9.6
Dawson Qld 40.01 49.6 9.6
Greenway NSW 38.65 49.7 11.0
Warringah NSW 38.71 49.7 11.0
Macarthur NSW 38.85 49.8 11.0
Bowman Qld 41.1 50.5 9.4
Dickson Qld 41.11 50.5 9.4
Aston Vic 36.85 50.5 13.7
North Sydney NSW 39.96 50.7 10.8
Hinkler Qld 41.66 51.0 9.3
Wannon Vic 37.63 51.1 13.5
Gilmore NSW 40.59 51.2 10.7
Flynn Qld 42.28 51.5 9.2
Petrie Qld 42.55 51.7 9.2
Casey Vic 38.65 51.9 13.3
Hughes NSW 41.45 52.0 10.5
Stirling WA 47.96 52.1 4.1
Flinders Vic 38.89 52.1 13.2
Longman Qld 43.25 52.3 9.0
Hasluck WA 48.18 52.3 4.1
Menzies Vic 39.33 52.4 13.1
Herbert Qld 43.76 52.7 9.0
Goldstein Vic 39.97 52.9 13.0
Sturt SA 43.2 53.0 9.8
Blair Qld 44.31 53.2 8.9
Kooyong Vic 40.42 53.3 12.9
Robertson NSW 43.13 53.3 10.2
Cowper NSW 43.25 53.4 10.2
Dunkley Vic 40.62 53.5 12.8
Paterson NSW 43.68 53.8 10.1
Higgins Vic 41.24 53.9 12.7
Boothby SA 44.63 54.2 9.6
Page NSW 44.54 54.5 9.9
Gippsland Vic 42.3 54.8 12.5
Dobell NSW 45.16 55.0 9.8
Bennelong NSW 45.87 55.6 9.7
Moreton Qld 47.17 55.6 8.4
McEwen Vic 43.58 55.8 12.2
La Trobe Vic 44.17 56.2 12.1
Eden-Monaro NSW 46.73 56.3 9.6
Lindsay NSW 47.08 56.6 9.5
Corangamite Vic 44.68 56.6 12.0
McMillan Vic 45.01 56.9 11.9
Wentworth NSW 47.49 56.9 9.4
Deakin Vic 45.03 56.9 11.9
Bonner Qld 49.49 57.5 8.1
Makin SA 49.07 57.9 8.8
Wakefield SA 49.33 58.1 8.8
Parramatta NSW 49.17 58.3 9.1
Kingston SA 49.93 58.6 8.7

Let me also provide evidence from the last three elections on the basis of the random distribution of voter change. If we run a simple scatter against the size of the swing in seats vs the margin those seats were held by, and a run a simple regression through them, we get:

98electionscatter1.jpg01electionscatter1.jpg04electionswingvmargin1.jpg

The Margin axis on the bottom is positive for Coalition seats and negative for ALP seats. So an ALP seat with a margin of 5% would be represented as -5 in the chart, and a Coalition seat on a 5% margin would be +5 in the chart. Likewise with the swing, a positive swing means a swing to the Coalition, and a negative swing is a swing to the ALP. The red line is the regression which helps show the general relationship between margins and swings. There is certainly lot’s of variance there – that’s why we adjusted our swings downward in the above estimates to accommodate for the general consequences of the variance at a state level. The scatters are also quite interesting in their own way.

AddThis Social Bookmark Button

32 Responses to “The Proportional Application”

  1. Mark said

    Brilliant Poss. Think you’ve just made the Mackerras pendulum obsolete.

  2. Stunkrat said

    Hmmm…North Sydney 10.8 – I suppose this confirms that Buffy is in some trouble?

  3. dany le roux said

    Possum,
    Wouldn’t it be for the first example ( 60/40 ALP/Co) if there were a 5% change there would be a gain of 2.5% for ALP and a subsequent loss of 2.5% for the Coalition? This would result in a new ratio of 62.5/37.5 showing a swing of 5%.
    Likewise the other two ratios viz 52.5/47.5 and 42.5/57.5.
    I think you have the correct result because the swing is actually 10% .
    I think a TPP of 10% means that 5 people in 100 have changed their minds or have I made a very big and fundamental mistake?

  4. Possum Comitatus said

    Danny, we’re working on a swing rather than a change.I didnt exactly word that well… oops.

    I’ll fix that up.

  5. Bushfire Bill said

    And I had always thought “swing” and “change” were synonymous.

    Possum you may find the scatters interesting, but could you let us in on the secret?

  6. habby said

    Love your work!

    My kids often say to me when I’m trying to explain something to them “Dad, can we have the short version please”.

    So for those of us who are time (and/or statistically) challenged a little one para summary or take home message would be great. (If for no other reason so I can decide whether I’ve undertood it all!)

  7. Dan said

    Poss – you’ve made my day. Will print this out and ruminate further this evening.

  8. monk said

    Labor Primary Support Remains Above 50%

    ——————————————————————————–
    Federal Poll : Finding No. 4219 : October 5, 2007
    On September 29/30, the ALP primary vote dropped 0.5% to 53.5%, while Coalition support also fell 0.5% to 35.5% during the same period. With preferences distributed as they were at the 2004 Federal election, the two-party preferred vote is ALP 61% (up 0.5%), L-NP 39% (down 0.5%). If the Federal election had been held last weekend the ALP would have won in a massive landslide, the latest ‘face-to-face’ Morgan Poll finds.

  9. alpal said

    Possum: Is it correct that Morgan FtF pick up a voting pattern before Neilsen and Newspoll – and that they then follow? Why are prefernces distributed the same as last election in Morgan polls? I’ll have some info on MacPherson on Monday – probably no different to GC Bulletin story.

  10. KC said

    Poss

    Brilliant, it lines up with individual seat polling, Morgan poll now showing 61-39, so confirms many are very pissed off.

    I like the individual results, you can factor in your own member and local issues swing to estimate whether the seat will go or hold.

    For example Wentworth, Turnbull has to rely on a 5% member margin to be safe, whilst Hockey looks in big trouble in North Sydney as he would have a negative local margin due to his ineptness. Boothby looks promising for Cornes as the negative press she has got would mean a ngative margin for her but she may still get up.

    And Howard is in big trouble in Bennelong, similar to Turnball.

  11. Samuel K said

    Hi Possum

    Thank you for your very interesting work – it all makes sense to me.

    I think your scatter plots are very telling in that they show just how silly it is to project individual seat results from predicted national or state swings. It’s hard to believe, but a significant number of seats actually swung to Labor in 2004 under Latham. And I believe that the baseball bat election even had a few seats go to Labor under Keating. I guess despite the expectation of a Labor win of significant proportions, we can only expect that there will be some leakage – with a few seats (perhaps in WA?) going to the Liberals in 2007 under Howard. In any case, we can expect some big variation. Your work is very interesting – but I fear it is a bit like all theoretical probability work – it is swamped by statistical noise in practice.

  12. Martin B said

    hi Poss

    Are you applying the average swing of a safe govt seat to all safe govt seats? If so, you are likely to be overestimating the swing in govt seats that are less safe than the average for this category (and underestimating the swing in seats that are more safe than the average for this category) as I suggested in my last post to the MWABSYH thread.

    Although I have to say the swing vs margin plots look pretty scattered to me (so that isn’t a lot of support for the SPoVS model🙂 ).

  13. I’m sorry, but those scatter plots are not at all impressive that there is a relationship between margin and swing size. Looking at the regression line for 2004, the swing against Labor was 2.5% in a Labor seat where the Labor 2PP was 70%, but 1.5% in a Coalition seat where the Labor 2PP was 30%. So if you more than double the Labor vote, the swing only increases from 1.5% to 2.5%? It may be statistically significant, but it is not politically significant. You’ve drawn a couple of straight lines with a slope that look important, but don’t actually turn out to be significant. If there was something in the theory that a seat with lots of voters for one party is more likely to swing against that party, the slopes in those graphs would have to be much steeper. There have been numerous papers comparing the idea of uniform swing in seats comparing it to the idea of proportional swing, where the swing size depends on the underlying vote, and the relationship is always weak.

  14. canberra boy said

    AC Nielsen has provided a six-month amalgamation of its monthly national polling for April to September to the SMH & The Age.

    It shows a 9.8% national swing, and State swings of 9.9% for NSW, 7.9% in Victoria, 12.3% in Queensland, 14.9% in SA & NT (combined), and 5% in Western Australia.

    The results seem to be pretty consistent (that is, within the margin of sampling error) with the most recent Newspoll quarterly figures, which showed 8.8% nationally, 9.2% in NSW, 11% in Victoria, 9.1% in Queensland, 9.4% in SA, and 4.4% in WA.

    The higher Nielsen figure for SA&NT may explain the recent Coalition panic about the seat of Grey.

  15. Possum Comitatus said

    CB at 13 – those ACN numbers do seem to be very consistent with Newspoll, and very consistent with individual seat polling.This isnt bullshit here – these are the actual swings in play *at the moment*.

    Marty at 12,
    Above, it was only state swings that were used.Once they were applied to each seat on the basis of the proportion of coalition voters in a seat, I could then tally up the averages for the three seat types and compare those results to the Newspoll results.

    KC and Samuel at 10 & 11,
    You’re right about the variance issue.The key with something like this is to look at the individual seat projections as the baseline figure, and from there, then apply political and demographic issues onto those individual base numbers.For instance, we know that the 25-39 age group is one of the biggest swinging demographics in the country, and we know that they are swinging largest in Melbourne and Sydney.SO we could take that into account when we look at a seat like, say, North Sydney which has the 7th highest proportion of 25-39 year olds in the country.That would explain why Hockey is in trouble – not only as a basic issue of the random distribution of coalition voters changing to the ALP, but also because of demographic issues like that.On the ohter hand, we can also look at Cowper only having half the proportion of 25-39 year olds that North Sydney does and reduce the size of the swing we could expect there.

    Likewise, with a seat like McPherson (which isnt mentioned but has a projection of 46.3% ALP TPP), we can then overlay the demographic issues in play that we discovered were operating in McPherson over here:
    https://possumcomitatus.wordpress.com/2007/08/30/hasta-la-vista-mcpherson/

    The base line level from which the variation in the swing launches off isnt a uniform swing in most elections, its a swing based on the random distribution of voter change.

    Alpal at 9,

    On a lot of issues they seem to.Interest rate rises and falls for instance is one example where the F2F results are easily measurable and statistically significant as being a leading indicator of movement in the next few ACN and Newspolls.

    I suppose it comes down to a bit of personal interpretation on which events you think would actually shift the 5% or so of newsvoters in the polls.F2F movement is usually noise like other polls, but there also seems to be a bit of first impression basis to the way its vote estimate changes.Over the last few months, nearly everytime that I think that the government has done something during the week which would turn people off – the F2F usually moves the ALP vote up.And F2F polling does tend to exacerbate that effect anyway because people are forced, in a sense, to answer questions in front of a stranger about issues that they may not have been thinking much about.So when push comes to shove, the most recent events tend to be the ones that are strongest in the memory from which to spout an opinion (for those voters that arent rusted on or absolutely adamant about which party they are going to vote for).

    Most pollsters actually distribute preferences according to the 2004 election.It comes back to the problem of the way people say they will preference other parties in a poll doesnt always stack up to the way they actually do on the ballot paper when they have a how-to-vote-card in their hand.Morgan on its web site also has a measure of the TPP for every poll based on how the voters say they will distribute preferences.That measure goves a slightly more favorable preference distribution to the ALP than the 2004 election – likewise, ACN polls also give a more favourable ALP preference distribution when they ask survey respondents about their second preferences.

    Thanks for the McPherson stuff.I’d heard Saroff was 2 points behind from a couple of different people, but that could just be an interpretation of the “couple of points behind” spiel which looks like the official ALP leak message.

    Habby at 6,
    That’s a good idea.I might try to give a brief conclusion at the end of these sorts of posts from now on.

    BB at 5,

    In the 98 election, you can see how the One Nation effect played out across all the seats – making the scatter very sparse, but with that tight cluster of government marginals around the 4% margin only swinging against the government by a small percentage (that was the One Nation preferecing of the COalition in the marginal seats at work).That same group of seats at the next election returned nearly all of those marginals back to around their 96 buffer levels (which is that cluster of dots around the 0-5% margin that swung to the Coalition by about 1-4%).Then in 2004, you had a lot of those same seats swinging to the government by up to 6%, and another chunk of those seats swinging up to 2% against the government.That sort of shows that these seats arent homogeneous on the issues that drove the 2004 election.So issues like the economy dont appear to be a strong enough resonator as an issue to support a government resurgence in enough marginal seats, should they choose to make the economy the key plank in their campaign strategy.

    To get an idea of how the 2007 election would play out at current polling levels, if you take the 98 election
    diagram and visually shift the whole thing down to where the dots are centered around -8,and the red regression line keeps its slope but crosses the -8.5 mark at about Margin value 0, that gives you an idea of the type of result.I would be surprised to see any seat swing to the government.

  16. canberra boy said

    I’ve played with a spreadsheet to amalgamate the two quarterly Newspoll reports for their polling over the April to September period with the Nielsen figures for the same period I quoted above. I weighted the swings by the relative size of the three sets of polls.

    I have to note some reservations – such as using Nielsen figures for combined SA&NT with Newspoll details for SA, and the possibility that the Newspoll amalgamations included the same July polling in both quarterly reports.

    Here then are the results of all Newspoll & Nielsen polling for the last 6 months: national swing 9.5% (ie implied 2pp of 56.7 to 43.3%), NSW 10.5%, Vic 9.1%, Qld 11.4%, SA 11.9%, WA 5.0%.

    I suspect that there may be statistical doubts about whether combined results of multiple polls have the same error margin as one big poll, but both Newspoll & Nielsen report their results as if that is the case. If so, the error margins for the above figures would be roughly: national 0.67%, NSW 1.2%, Vic 1.3%, Qld 1.6%, SA 2.3%, WA 2.2%.

    When we compare these figures to the Newspoll July to Sept figures which Possum has used for his most recent work, the most interesting points are that the combined results show the swing in SA to be 2.5% more, Qld 2.3% more, and Vic 1.9% less. It may help to explain some of the recent stories about who’s concerned about which safe seats.

  17. Possum Comitatus said

    Thanks for that CB, it also highlights that even in the worse case scenario – these swings aren’t some statistical trickery or overblown results.These are great whopping swings that stand on their own merits.

    The SA results, because there are so few seats in SA, only require a couple of seats to be acting strangely and that would easily explain the variance.If Grey really was revolting, that would do it – likewise if one of the more likely ALP safer seat gains of Sturt or Boothby are swinging a lot less, that would also explain a few things.

    Victoria though, by way of having so many seats (and hence each seat having a smaller influence on the overall state-wide results) still has me baffled. Qld is becoming more explainable as a result of the big swings in McPherson and largish but not enormous swings in Dawson, Fairfax and Leichhardt that are apparently happening.Forde is not behaving well for the Coalition, especially being a three cornered contest – that’s why Howard was out there this week.

  18. I have to say I don’t find the straight line best fits on the swing graphs at all persuasive on the argument about swing by safeness. From the 2004 graph, the swing to the Coalition in 70% Labor seats turns out at 2.5%, but 1.5% in 30% Labor seats. That is not much of an increase for a doubling in size. There have been several papers over the years looking at whether swing is uniform or proportional to existing vote, and while there are cases where swing is bigger in safe seats (eg in safe Labor seats in 1987), uniform swing fits the model better. The straight lines might be statistically significant, but they are not politically significant.

    Remember that safe seats have a bigger base of voters rusted to one side. That’s why they are safe. Marginal seats are marginal because the core of rusted on voters for each side is roughly equal, which means the

  19. Possum Comitatus said

    Antony – Askimet doesnt seem to like you much, you’re ending up in bin with the insurance sellers and penis enlargement people.Oh the inhumanity!🙂

    The statistical significance is all that’s required here.The political and demographic issues for individual seat vote levels obviously get played out in indiviual seats/areas/states etc, but it’s what is beyond that variation that I’m looking at here.

    The first difference of the regression in those three examples are not large, but it’s large enough to make a
    difference with a coefficient of the margin for the 98 election being -0.05 with a Std Error of 0.02 and significance at the 2% level.Likewise for 2001 the coefficient on the margin was -0.045 with a Std Error of 0.017 and significance at the 1% level.The relationship is only weak in magnitude for seats on small margins.The greater the margin the stronger the effect plays out.

    At 2004 the variance swamped the swing, making the the margin coefficient statistically insignificant.In 96 what was interesting is that there actually was a uniform swing.

    So for some elections the swing is uniform, for some not.The polling data, the only observable quantitivate data we have on the current state of play suggests that this is another non-uniform election like 98 and 01, being a proportional vote change basis instead of a uniform swing basis as the most accurate representation of the underlying data generation process from which seat by seat variability then applies from.

    In all three elections, the seats with the large margins tended to be dominated in their swing variation by
    a random distribution of voter change.In 98, the coalition seats held by more than 15% were dominated in
    their variation from the uniform swing by a larger number of those seats swinging greater than uniform.The same thing happened for the Coalition in 2001 (on the downside) and 2004 was all over the place.Similarly for the ALP bigger margin seats, they had smaller swings toward them in 98, and larger swings against them in 2001 and 2004.

    Across all three elections, those seats with larger margins acted more like a random distribution of voter change than they did as a uniform swing.

    On the issue of safe seats having a greater number of rusted on voters.That is probably true, we only have partial estimates of what that level would be (little more than a guess really), and we have no data on the threshold of a ‘rusted on voter’.We dont know the point at which a presumed rusted on voter would tip to actually voting for the other side, but we know it does happen in reality, we see it in places like Qld under Beattie.What were once presumed rusted ons voted ALP.

    As a result, and taking this from a pure data perspective, there is more evidence to suggest that probability theory utilising a random distribution of voter change is a more accurate representation of the underlying data generation process (from which seat by seat variability then becomes applied) than a uniform swing assumption is for this election.

    There may be even better explanations of which swing type is the most accurate representation of the underlying data generation process than either of these.I’m playing around with more complicated versions at the moment to see what happens.The marginal seats do tend to take the weight, but that also probably has to do with marginal seat campaigning.

  20. [you were in the spam bin again]

    Your second last paragraph is nonsense Possum. Your argument against a uniform swing is setting up a straw man, as no-one seriously argues that the swing is ever uniform, in the sense of being the same everywhere. The swing is never uniform like that, but instead distributed around the average swing. The question is always about the standard deviation of the swing, and also as to whether there are factors that cause the swing to vary in some seats compared to others.

    What matters is the absolute size of the national swing, and the standard deviation of the swings by seat. In 2004, the Standrd deviation on the swing was 2.8%, less when you account for state differences, and even less if you take into account outer suburban seats that swung against Labor.

    If the same standard deviation were to apply in 2007, then if the national swing was only 5%, that would matter as the government would have a chance of hanging on to marginal seats. But if the swing was 7%, it would be irrelevant as the government would have little chance of holding on to enough seats given the standard deviation of the swing. Given a swing, a standard deviation and knowledge of the margins, it’s a simple probability model from there.

    We use a regression model on election night that does exactly that, with factors built in for safe seats, region, state and margin. I have to say, it really isn’t much advance on just assuming a uniform swing and banking on the liklihood that seats that don’t fall on a uniform swing, or fall on a swing beyond the uniform, generally cancel out.

    From years of preparing rehearsal data, why you try and factor in seat by seat trends, I have to say that the swing on election night is always more uniform than that produced by trying to factor in local factors.

  21. I wonder if the underlying behaviour of these 6%-15% safe Lib seats is more like what the chemists call a buffered system. That is, where the underlying dynamics resist new influences up to a point, and then a threshold is reached and a wholesale change takes place. I suspect that this might mean that most of these seats will actually stop well short of the swings in your list whilst a minority will go off into uncharted territory with swings above 10%. Those in that minority might well be in innerish Melbourne and Sydney where boomers with small l values have changed the demographics.

    If the last poll before election day is the same as the current crop, I expect some surprising Lib loses on election night, but I also expect the majority of your list of Lib seats above 6% margin will survive. That would give us a Labour tally in the neighbourhood of 80 something rather than 100ish as indicated by your calcs. Presumably this is where harder heads in the Liberal machine are feverishly attempting to build a firewall. I hope they fail miserably.

  22. Possum Comitatus said

    Antony,

    The reason that the uniform vs random distribution argument is far from a strawman is simple:

    We don’t have the standard deviation of the results for the 2007 election.

    I agree with absolutely everything you’ve said here about managing the variation of the swing… but as a rear view mirror process only. It can only happen once the actual election data starts coming in, not before…. Obviously.

    So without having the luxury of being a successful fortune teller, when we are trying to give meaning to what these aggregated poll results look like today, we really have only two choices.

    We can apply those state swings as a uniform swing to get a rough baseline look at the seats that are likely to be in danger, and then argue over local factors, demographics and individual seat polling to obtain a better picture, or we can apply those swings as a random distribution of voter change and then argue over local factors, demographics and internal polling results to obtain a better picture.

    At the moment, the observable quantitative evidence overwhelmingly supports the use of the latter over the former as the most accurate representation of the underlying data generation process that we can, AT THE MOMENT, observe.

    Alternatively, we could all just say ‘what’s the point – these are the swings’ and leave it at that. But if we just did that, we never would have figured out that seats like Grey, McPherson, Leichardt and Forde for instance were more likely than not to be having fairly large swings against them… which the internal polling subsequently verified after we’d identified them as possibilities.

  23. I’ll safely predict Grey, Leichhardt and Forde will have above average swings. Each has a retiring MP. I don’t need a break down of Newspoll to know that. Many of the seats you name from your models as likely to swing are fairly obvious if you look at history, or you look at where there are substantial discrepancies between state and Federal voting. Though I’m with Adam Carr in thinking there are a couple which look way off beam. As for not knowing the standard deviation, I can assure from years of experience, it is relatively stable over time.

  24. fred said

    “…which the internal polling subsequently verified after we’d identified them as possibilities.”
    Thats the key point.
    Prediction verified.
    Sort of.

  25. Possum,

    Very much enjoy the site.

    Re this post: are you suggesting that parties always do better in their opponent’s heartland, or merely that this is a feature of the Howard government?

    If you’re suggesting it as a general rule, then I’m inclined to agree with Antony. It’s not. The most obvious recent example is Victoria 2006, where Labor under Bracks performed better in its own heartland than in its Liberal/National opponents’:

    Moreover, if parties always did better in their opponent’s heartland, we would expect to see a convergence of margins over time, which we don’t.

    If you’re just suggesting the Howard government has done better in Labor heartland, and that it looks set to do so again in 2007 (albeit in the context of a big swing against) then of course you’re right.

    Regards,

    Nathan.

  26. steve_e said

    Possum, as an exercise I have applied the potential gains in seats from the above table to the Electorate Maps. This will make it easier for me to understand the TV coverage on election night.

    If the implied gains to ALP occur in the Melbourne suburbs the only retained LIB seat is Higgins.

    There is a sweep of 5 seats from Wannon in the South West along the coast to Gippsland in the far East. In regional Victoria, the LIB seats retained would only be Indi and Murray (both south of the Murray river). The NATS retain Mallee (South of Mildura). This is a total of 4 Coalition seats from a current holding of 16 being 7 Regional and 9 Suburban seats.

    In Sydney, the positon is an implied loss of Lindsay, Parramatta, Bennelong, North Sydney and Wentworth near the Harbour for 5 suburban seats. South of Sydney implied seat changes include Macarthuer and Hughes (Cook is retained to LIBS). North of Sydney changes include Robertson and Dobell. The other retained LIB seats are North of the Harbour (Greenaway, Mitchell, Berowra, Bradfield, Mackeller and Warringah) or 6 in total from a previous 16.

    Regional seat changes total 4 (Page, Cowper, Paterson and Eden-Monaro all with a coastal edge).

    In South Australia it is even more significant if control of Grey changes due to the retirement of the sitting member (this is the area North the Gulf of St Vincent). The only retained LIB seats would be Mayo and Barker all the others would become ALP seats.

    In Tasmania, the 2 LIB seats are expected to change to the ALP resulting in all 5 seats being ALP controlled.

    The change in the Electoral Maps in these states is most dramatic and shows just how hard it will be for the LIBS to recover from the expected outcome of the 2007 Election.

  27. Possum Comitatus said

    Antony (you were in the spambin again),
    So far, all we can say is that two separate approaches have lead to the same conclusions for a number of seats.

    There will be some seats that are way off beam, there are always seats that defy the swing.Unless some particular demographic issue, or some measurable economic effect is playing out in those seats which cause that swing defiance – then its not an issue I deal with, its an issue for the political scientists.

    The Standard deviation could be a constant as far as it matters in this circumstance – it still doesn’t help us much *_at this stage_* of the electoral cycle.At this point in time, it can at best provide us with information necessary to undertake a couple of dozen theoretical probability distributions from which to simulate (monte carlo or otherwise) the individual seat votes based on polling data and swing anchoring.But that still wouldn’t actually provide us with any new information – derived or otherwise, it would just be a more sophisticated, but not necessarily more accurate guess.

    That’s really all I can think of as far as a use for a consistent standard deviation in the swing at this stage, but if you have any suggestions, I’m all ears.

  28. Possum Comitatus said

    Nathan,

    I was suggesting it for the period of 98-04 federally.That’s the way it seems to have worked for Howard, both in the 01 and 04 elections where the swing was too him, but importantly in the 98 where the swing was away from him.When you look at the basic political modus operandi that Howard has used since 1996, the “If the low and middle income voters are looked after and the Coalition has a higher than average proportion of the 55+ vote, it’s nearly impossible to lose” strategy – you can see why it turns out that way.I’d also argue that when the electoral tide turns against him, there will be Coalition blowback from that demographic targeting strategy.Those low and middle income workers will flow back, resulting in a pattern scatter similarly shaped to the 98 election outcome, but where it wont be the result of One Nation,and where the swing is much bigger.

    Some of those State results are give a really good insight into the way the State election strategies worked out on the ground.Because the State electorates seem to much more demographically concentrated than the larger, more diverse federal electorates, elections where governments try and consolidate a win, or where they attempt to shore up marginals… or even campaigns where the opposition takes a two term strategy of trying to shore up their traditional supporters seem to play out well visually in the resultant margin/swing scatter.Some of the Qld results are pretty interesting in that respect (like the last Vic election looks to be from that spiffy chart of yours), especially when you use a ‘nearest neighbour’ line fit or some nonparametric regression line fit rather than an ordinary least squares job.

  29. Possum Comitatus said

    Robert at 21,

    That’s a good question, would the demographic characteristics of some seats be such that if issue positioning continued to run in the ALPs favour, would a tipping point be reached where a breat big swag of voters all came across at once, blowing out a swing rather than there being a simple slow, incremental movement to the ALP over time? I suspect that in some seats that would be the case – especially over particular issues….Childcare comes to mind.

    Those small ‘L’ liberal voters you mention might too if Howard tries to shore up the One Nation part of his vote (to save regional Qld seats), only to have it blow back in his face in more inner urban, affluent seats.

    The Newspoll safe government seat estimated swing is probably highly influenced by those seats that were usually marginal, but now hang around the 6-9% margin.One would expect to see 12% swings in some of those seats simply as a result of the demographics Howard has always based his political strategy on, turning against him.If you artificially hold up your political support in a demographic with fairly shallow measures (like scare campaigns and a bit of bribery) over a number of elections, the margins in those electorates will be plumper than they would naturally be, meaning that if the key demographics that make them up turn against the government, the swing against the government would likewise be expected to be plumper than it ordinarily would be.

  30. Possum Comitatus said

    Steve E,

    The Libs must be praying that the variation in the swing within States is high, and that the composition of that variation is such that only a handful of seats are taking most of the swing weight.. simply to minimise the bloodbath.I’ts not much for them to hold on to, but at least it’s something.

  31. Martin B said

    As for not knowing the standard deviation, I can assure from years of experience, it is relatively stable over time.

    So, er, what is the standard deviation of the standard deviations?🙂

    Antony, without giving away any patented secrets🙂 are there in your opinion any significant factors that make one seat more variable than another? You have mentioned retiring members but have suggested that other local factors are less important.

    The component of the swing (to ALP) proportional to the number of (Coalition) voters appears to be real, but small.

    Are there any demographic factors that make a seat likely to be more swinging than another?
    Or is it just esentially random behaviour from seat to seat?

  32. Preslar said

    Once again the 50 inch lg tv is one of the hottest gifts this Christmas, they’re selling out fast. I found mine at http://astore.amazon.com/50-inch-lg-hdtv-for-sale-20

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
%d bloggers like this: