Polity by Rob Salmond

Read Post

Polity: Meet the middle

254 Responses

First ←Older Page 1 7 8 9 10 11 Newer→ Last

  • Joe Wylie, in reply to David Hood,

    I spent entirely too long this evening looking up what colour election orange was. I am, for the record, now using #f89828 for graphs about the election.

    Sounds like one of those food colouring agents that was banned long ago in first-world countries on suspicion of triggering seizures.

    flat earth • Since Jan 2007 • 4593 posts Report

  • BenWilson, in reply to David Hood,

    As I understand it the left right metaphor is one of strategic positioning (hence the move to to middle arguments) rather than a multidimensional strength of signal metaphor.

    OK, but I'm not seeing that last graphic undermining the strategic positioning argument. It looks like there's a big chunk of people between the parties that National contested better. If I can see anything it that graph it's almost confirming Rob's point. The median is closer to National, so that's where the median voters went.

    That said, I don't really get how you made that graph, why you scaled up massively. There's some math about scaling between 2 fixed points that you haven't explained.

    Auckland • Since Nov 2006 • 10657 posts Report

  • David Hood, in reply to BenWilson,

    There are up to ten points of steps and I wanted something that divided evenly by 10, 9, 8,7 etc- I could also have gone to decimals.

    Dunedin • Since May 2007 • 1445 posts Report

  • David Hood,

    I think the graph is fundamentally misleading in that it is based solely on the Labour National dimension, and winning the median of that means a party is pretty closed to being able to govern by themselves- it completely ignores coalitions.

    Dunedin • Since May 2007 • 1445 posts Report

  • BenWilson, in reply to David Hood,

    Attachment

    I could also have gone to decimals.

    Ah, that explains a lot.

    I think the graph is fundamentally misleading in that it is based solely on the Labour National dimension, and winning the median of that means a party is pretty closed to being able to govern by themselves- it completely ignores coalitions.

    Yes, it’s an approximation with a lot of bias. There are so many ways in which the basic assumptions of the median voter theorem are violated. The multidimensionality is the biggest first point, especially in NZ under MMP. But also the theorem assumes unimodality and I think it’s pretty clear from the way voters placed themselves that unimodality is highly questionable. The theorem also assumes all voters vote (which we know to be false), and that they vote for their preferences (rather than tactically). I think the subsetted PCA here shows that that is very much questionable in the case of NZF. It looks like they actually place themselves closer to Labour. This shows that the left-right placement is not the dominant factor in their political choice, despite it being pretty clear that NZF voters are close to the LR median.

    In other words, it looks like LR matters most to people who have more extreme LR. Voters in the center don’t care about it so much, so LR moves won’t be what captures them.

    Auckland • Since Nov 2006 • 10657 posts Report

  • BenWilson,

    Looking more closely at the subsetting there, it does seem to divide Labour into 2 clusters. The dense cloud on the left near where they think National is, and the big sparse cloud on the right (the dimensions are reversed on that graph, it seems), in which their variable loading falls. Not sure what to make of that.

    Auckland • Since Nov 2006 • 10657 posts Report

  • David Hood,

    The unmorality of the idea seems to be getting a lot of criticism in its application to US politics at the moment, because it is so polarised the "median" is a pretty empty space

    Dunedin • Since May 2007 • 1445 posts Report

  • BenWilson, in reply to David Hood,

    because it is so polarised the “median” is a pretty empty space

    It looks to me like LR is trimodal. In more dimensions that might actually break down, resembling a normal distribution more and more (obviously I mean the higher dimensional equivalent). That's what I'd expect...but I might be wrong.

    Auckland • Since Nov 2006 • 10657 posts Report

  • David Hood,

    Dunedin • Since May 2007 • 1445 posts Report

  • Russell Brown,

    A pretty amazing longread on Corbynmania by Taylor Parkes for The Quietus.

    Parkes argues (and he's hardly the only one) that it's not Corbyn's policies that will spell disaster for UK Labour, but the company he keeps and the things he is on record as saying.

    The most interesting observations are around obsessive anti-imperialism.

    Auckland • Since Nov 2006 • 22850 posts Report

  • Steve Barnes,

    Peria • Since Dec 2006 • 5521 posts Report

  • Steve Barnes,

    Perhaps there is no Centre...
    I was wondering how much language has to do with the framing of the concept of Left v Right.
    For instance. You can say, "right = right and if you are not right then you are left with the left.". This gives the impression that left is second best at best.
    Alternatively, if you accept that doing the right thing for people and planet, without a self serving agenda, if you have empathy for others, if you believe in justice and fair pay, that you accept that there are people out there who are prepared to devote their lives to hold society to those values and that you have the right to vote those people into power, then, surely, that is the ideal.
    You can move away from that ideal in several ways, Tyranny, Fascism, the Law of the Jungle, whatever.
    The ideal is the centre, all the others are distractions for the benefit of the few.
    So, it is only right that the centre is Left.
    ;-)

    Peria • Since Dec 2006 • 5521 posts Report

  • BenWilson, in reply to Steve Barnes,

    Attachment

    Something I’ve had to put on hold while getting other things done. Using opinion on subjects rather than about left-right positioning, to look for principle components. I had someone do this in SAS for me, I’ll attempt to do the same thing in R when I get the time.

    The party points are the center of mass for all the people that voted for them. The other labels are the variable loading positions for the entire population on the questions that they correspond to. You’ll have to ask me which actual questions they were if you're interested in one particular question.

    But the thing that leaps out at me from this graph is that while of course we know that politics is multidimensional (or at least anyone who knows what that means probably thinks that), in this graph Labour and National are sitting almost squarely on the principle axis, and 90% of the variation in the data is explained by this axis.

    So if we did actually like Rob’s definition of left and right as being the direction between Labour and National, then just on opinions about politics, as asked in 2011 in this survey, it does look like we’re almost in a uni-dimensional situation. I didn’t go off every question, there’s a lot more tidying to do on this…but it’s bloody interesting. These were the general questions you’d ask someone to find out whether they’re left or right wing without just asking them if they’re left or right wing.

    This was built on that helpful someone I don’t want to name without their consent giving the suggestion to binarize the choices, because unfortunately the data allowed “Don’t know” on the same continuum as the answers and I lose almost all the data if I keep only complete cases. It seemed like a good alternative to imputing the don’t knows.

    Auckland • Since Nov 2006 • 10657 posts Report

  • Steve Barnes,

    Nice to see Lazyben is so popular amongst National supporters. ;-)
    As for the "don't knows".
    These are the people that are generally targeted by the right, the "Undecided Voter", someone who, obviously, reads little that he doesn't agree with, believes in his own opinion wholeheartedly and that belief is that the left are loony because it sounds right.

    Interesting work Ben. That National and Labour seem to be on the same plain. I was thinking that if that plain was a brane the 3D modeling could be quite enlightening.

    Peria • Since Dec 2006 • 5521 posts Report

  • Rich of Observationz, in reply to BenWilson,

    By principle you mean principal as in "the principal component" rather than principle as in "the principles of the party"?

    Back in Wellington • Since Nov 2006 • 5550 posts Report

  • BenWilson, in reply to Steve Barnes,

    Interesting work Ben. That National and Labour seem to be on the same plain.

    Thank Steve, although it's far from being as complete as I'd like. I'll be fleshing it out a lot more over the next month.

    As for the same plane thing, well the analysis itself looks for the plane (well, hyperplane, technically) that most separates the data out. Since most people voted for either National or Labour, I'd expect them to have ended up along the first principle axis somewhere. Essentially, it was looking for the plane that would go between them, rather than luckily finding one. I did not tell the analysis anything about the left-rightness of the questions, nor how the people voted. It just looked for the best way to split them into two groups based on all the questions. Then it does it again for the second dimension, and so on. There's more dimensions, but I only plotted the first 2.

    There are some interesting things on there. taxsolve is quite a long way from ACT and National. This is "Reducing taxes in general would help solve our economic problems". immig is a long way from NZ First and actually from everywhere, but this is because it was worded in the reverse direction "Do you think the number of immigrants allowed into New Zealand nowadays should be" with increased score meaning "increased". ie you'd expect NZF supporters to disagree with this. So the second axis (up-down) is most strongly associated with views on immigration, it seems. The strange location of taxsolve I can't explain.

    Auckland • Since Nov 2006 • 10657 posts Report

  • BenWilson, in reply to Rich of Observationz,

    Yes, principal components. But you knew that, I'm sure, you pendantic sew and sew.

    Auckland • Since Nov 2006 • 10657 posts Report

  • Rich of Observationz, in reply to BenWilson,

    “Reducing taxes in general would help solve our economic problems”.

    If you feed that sentence through a reading grader, it comes out at a (US) grade 12 (as in 6th/7th form - 17/18).

    That's a higher level of literacy than most people can manage, so it'd quite possible that the anomalous results are caused by a failure to understand the question.

    It's also a loaded question, because it implies its own correctness.

    Back in Wellington • Since Nov 2006 • 5550 posts Report

  • BenWilson, in reply to BenWilson,

    The strange location of taxsolve I can’t explain

    Well, OK, I can try. It's something that isn't important in the Labour/National distinction, but is important in the second dimension. In other words, it's not something that could distinguish Labour voters from National ones, but could help you tell NZF ones apart from Green ones, since they're very close in the first dimension. That and views on immigration.

    Obviously there aren't any questions about environmentalism here otherwise there would be some that were near the Greens. This is not deliberate in that I don't think it's important - I just didn't use the B1-5 questions in the survey yet because dealing with the "don't knows" is more problematic. I'll get around to it at some point.

    So I guess the point I made above about explaining 90% of the variation should be made with this in mind. It tells you as much about the questions as about general positioning. Had all the questions been about immigration then NZF and Maori Party would have ended up on the main axis at one end, UF and Greens on the other, and National and Labour close together in the middle, with Labour probably closer to the NZF end.

    Auckland • Since Nov 2006 • 10657 posts Report

  • BenWilson, in reply to Rich of Observationz,

    That’s a higher level of literacy than most people can manage, so it’d quite possible that the anomalous results are caused by a failure to understand the question.

    Could be. Or maybe it isn't really that much of a distinction between the main parties' voters.. Can dig deeper here. Won't be until the end of next week, though.

    Auckland • Since Nov 2006 • 10657 posts Report

  • Steve Barnes, in reply to BenWilson,

    Can dig deeper here. Won't be until the end of next week, though.

    Waiting with bated breath.
    Will we be going 3D?
    ;-)
    We could end up with a HollowGraph...
    /coat...

    Peria • Since Dec 2006 • 5521 posts Report

  • BenWilson, in reply to Steve Barnes,

    Will we be going 3D?

    I've been looking at it like that on my workstation, but I don't know how to give that out on the web. It would need some kind of java applet or something...

    I could show a 3D model in 2D if I could rotate it to find something useful to look at. But essentially that's what PCA is doing anyway - you're looking at this data rotated for maximal interest and then projected onto 2D. In this case, 95% of what is interesting is actually there in those 2 dimensions anyway - a third dimension might add like 2% to it. This is the whole point of PCA. It's a way of reducing dimensions so that we can look at them. The data underlying the graph has as many dimensions as there are variables shown there. The analysis just turned the whole thing around in every dimension until the data looked nice and spread out, so you can look at it (it also scales and centers it)

    It would be fun to rotate it around, but ultimately not particularly illuminating. This doesn't alway happen though. I'll be sure to show you anything I do find that's cool in 3D. I expect when I add in questions that are more differentiating for the small parties that less of the information would be captured in the first dimension here.

    That's really where I'm going with this. It's kind of cool that you can pretty much gauge left-vs-right on the subset of questions here, so long as you go with Rob's view that left and right are defined relative to the two major parties. I think it's also somewhat illuminating, that the right here are mostly the beneficiary and union bashing questions rather than broader questions of economic organization or management. That's possibly an actual insight, at least for Labour, and for all of us trying to comprehend what moving to the right actually means. It means, quite literally, moving toward those variables which (confusingly) are shown on the left, clustered around National and ACT.

    I'm not convinced, looking at this, that the pickings in that direction are really that rich for Labour. But this graph is, once again, very much a subset of political opinion questions. I want to get to a situation where most of the variables are clustered around most of the votes. This graph is not like that, and if one didn't know that National is dominant, one might think the majority of NZers were voting left.

    What that shows is that the questions here are not adequate to explain their voting behaviour. Many have said that valence issues are bigger than policy and positioning ones...I'd like to be able to confirm that, when I get the time. Gut feeling is that it's true.

    Auckland • Since Nov 2006 • 10657 posts Report

  • Rich of Observationz,

    I'd actually like to get the answers for more meta-questions, like:
    - have you always voted for the same party?
    - how do your parents vote?
    - when do you make up your mind on how to vote?

    Also, maybe a longitudinal study on how attitudes shift during elections?

    Back in Wellington • Since Nov 2006 • 5550 posts Report

  • BenWilson, in reply to Rich of Observationz,

    There's tons of interesting stuff to dig out of those surveys. The parental voting and when people made up their minds is certainly covered.

    Studying the voting across time is done to some extent, both in questions about how people voted last time, and also in actual comparisons for individuals that span multiple surveys. That's longitudinal, but I don't know how far it goes back.

    I'm quite interested also in questions about how people behave when they change their vote, either to another party or to not-voting. So many questions, so little time.

    You've got the skills, surely, Rich. Have a dig, see what you find.

    Auckland • Since Nov 2006 • 10657 posts Report

  • Steve Barnes, in reply to BenWilson,

    but I don't know how to give that out on the web. It would need some kind of java applet or something...

    This may be the go, if you're using SAS...

    Peria • Since Dec 2006 • 5521 posts Report

First ←Older Page 1 7 8 9 10 11 Newer→ Last

Post your response…

This topic is closed.