Some of my half-dozen readers may have seen this piece by Peter Kellner in last Sunday's Sunday Times, referring in part to an opinion poll which YouGov did at the beginning of March. The aim of this was to apply the process used in my on-line political survey to a balanced sample of the British population, to discover the axes which best describe Britons' political beliefs.
YouGov have very kindly allowed me to make use of the data from the poll. (Many thanks also to Tom Steinberg, who collaborated in this work.) Hopefully this is the sort of thing which some of you find interesting; there's quite a lot to say, and this is just a very brief summary.
To start with, you can answer the questions in the poll at, http://www.politicalsurvey2005.com/ and see how your views compare with the rest of the British population. Then read on....
As before, the idea here is to present respondents with a bunch of statements -- we used 32 in total -- and ask them the extent to which they agree or disagree with them. Some examples:
- `I am comfortable with the way that genetic engineering is being used in the food industry';
- `To get extra police on the beat, we should allow more suitably trained foreign police officers to move to the UK';
- `Prisons are too soft on criminals'.
Now we can look for the combinations of statements which maximise the variance in the population. These combinations correspond to axes of belief along which the population is most spread out. Once we have the axes, people's political views can be summarised by numbers which give their positions on those axes, and compared to the views of other members of the population.
So, how does this turn out? Principal Components Analysis gives us as many axes as we put in variables -- in this case, respondents' answers to the questions. But only some of these axes are statistically significant; in this case we find two, of which the first is very significant and the second of only marginal significance.
The most important ten statements on the first axis are these:
-- a person who agrees or disagrees with each of those statements in the sense given will end up at one end of the axis (we, arbitrarily, made this the positive end); a person whose views are the opposite will end up at the other end. Slightly facetiously we refer to this as the `Axis of UKIP': the extremal positive views are those of people who are Eurosceptic, believe in capital punishment and harsh prison régimes, and oppose immigration. At the other end of the axis people believe in further European integration, the primacy of international law, the benefits of immigration, and so forth.
This axis is identifiably left/right -- people at large positive positions are definitely `right wing' -- but it is not the traditional left/right axis of economics and class division.
Here's how the British population as a whole is distributed on this first axis:
(The thicker line is the cumulative distribution, measured against the vertical axis; read it as `the fraction of people to the left of this point'. The people named at each end of the axis are plausible stereotypes to whom you might ascribe the extremal views on that axis, but note that I'm not making a statement about their actual positions!)
Note that the tail of more internationalist people on the left is much bigger than the tail of more isolationist people on the right. I'm not sure why that is, but it may be because to push yourself very far to the right you have to agree to quite a lot of quite offensive things. And perhaps rehabilitationist/internationalist extremism looks less offensive than rightish extremism anyway.
Here's how people who said -- in early March, but the polls haven't moved that much since then -- they'd vote for each of the major parties show up on the first axis, compared to the whole population:
- Labour are left-of-centre;
- the Tories are right-of-centre;
- the Lib Dems are more spread out than Labour voters, but basically overlap completely with them;
- in the center, voters for all three parties are mixed up together.
On values, everyone is a swing voter now.
If it's the axis of UKIP, you might ask, where is UKIP? Here:
Observe that the median UKIP voter is in almost exactly the same place as the median Tory. Core UKIP voters look just like core Tory voters, and they have the same beliefs about Europe, capital punishment and immigration which Tory voters do. And, with curious symmetry, Welsh and Scottish Nationalist voters show roughly the same relationship to Labour voters as do UKIP to Tories:
Greens and non-UKIP nutters -- black is BNP, purple Veritas, and pinkish Respect -- look like this: (but please note that there are very few voters for these parties, so these plots are pretty approximate)
Now for the second axis. This axis is only marginally significant, but I'm going to talk about it anyway, because it makes sense conceptually. Here are the top ten statements:
Again, this is identifiably left/right in some sense -- but it is the right of the free market, free trade, and the war against Iraq. In summary, the person on the extreme positive end of this axis is the person who has read every Economist editorial of the past five years and believed all of them. The stereotype I would suggest here is Margaret Thatcher -- or Oliver Kamm.
Here's how the parties turn out on the second axis:
Note that Lib Dem voters are significantly to the left of Labour voters on this axis. Partly this is because of opposition to the war against Iraq; partly this is because the Liberal Democratic party no longer resembles is classical Liberal antecedents.
On this axis, Greens, Scottish and Welsh nationalists (light blue) and nutters look basically like Lib Dems:
-- though note a UKIP and BNP tail to the free-market right. (Again note that there are rather few supporters of those parties, so take those distributions with a pinch of salt.)
I'm going to leave this for now, otherwise I'll still be writing this while everyone else is in the pub. But those who are after further insights may find these slides from a talk I gave on this interesting.
Oh, and I should probably also thank the nice people at NTK for their helpful career advice.







Comments
Posted by Robin Grant, Friday, 15 April 2005 19:29 (link):
We're trying to collate everyone's answers here. Once you've taken the test, please pop by and add a link to your result.
Posted by Roy Badami, Saturday, 16 April 2005 01:39 (link):
Nice to see graphs confirming what we all already knew, namely that LibDems are left of Labour...
I've been trying to figure out what my politics are (in terms of categorisation) and while I tend to describe myself as a socialist, I don't think I really am. I don't believe in nationalisation of all industry, nor in an end to the market economy, which AIUI are traditional beliefs of socialists...
I am in favour of redistribution of wealth, and of a heavily regulated market economy. I think that makes me a social democrat (possible slightly towards the left of social democrats) but I'm not really sure...
All these terms are so fluid that it's difficult to find workable definitions of them...
-roy
Posted by soru, Wednesday, 20 April 2005 00:35 (link):
-- Nice to see graphs confirming what we all already knew, namely that LibDems are left of Labour... --
Surely you mean that lib dem _voters_ are to left of labour voters? Not at all the same thing.
Also, what the hell is 'Iraq' doing on the 'economic' axis? Probably should either ask another 10 questions to tease out a position on use of force/foreign policy, or just dump the question.
soru
Posted by Matt Daws, Saturday, 16 April 2005 10:52 (link):
Chris, Great stuff! But could you explain how you weighted the "agree" vs "strongly agree" etc.? When I did the test, I came out as very left-wing in the Axis of UKIP, despite mainly being vague (that is, not putting strongly agree or disagree to very much).
Posted by Chris Lightfoot, Saturday, 16 April 2005 16:25 (link):
We start with each respondent's answer to each question, and turn it into a variable like this:
(Actually it doesn't matter how you asign these scores -- using {-4, -1, 0, +1, +4} or {-1, -1, 0, +1, +1} gives basically the same axes. Not surprisingly the difference between `tend to' and `strongly' agree doesn't carry much information.)
Then we normalise everything to unit mean and unit variance (on the assumption, a priori, that no question is more or less important than any other). This means that you can get quite a significantly non-zero score even by saying `no opinion' to everything -- making apathy a political act, in some sense.
Then we compute the variance-covariance matrix among all the variables and compute its eigenvalues and eigenvectors (classical PCA). Actually there are lots of things we can adjust here, for instance using correlations rather than covariances and so forth, but it doesn't make much difference to the final result. MDS gives rather similar axes too, by a somewhat different method. I looked briefly at independent components analysis (looking for the axes along which the data form a non-Gaussian distribution), but it didn't produce any new insights.
Posted by Roy Badami, Thursday, 21 April 2005 19:51 (link):
I quite like the (-1, -1, 0, 1, 1) approach, because I worry that the dividing line between 'tend to agree' and 'strongly disagree' is too subjective to carry useful information. I suspect that's why you don't find it particulalry significant. If it were possible to construct a test that could objectively record the strength of opinion, then I imagine this distinction _would_ carry useful information.
Another interesting approach is to omit the 'neutral/don't know' answer from the questionaire, and then score with (-1, -1, 1, 1). This is the scoring used in Simon Baron-Cohen's AQ test for Asperger's Syndrome.
Arguably this makes the questionaire far less subjective in its interpretation by the respondant, though I've no idea how well it would work for something like this. I suspect that 'neutral' carries important information here that you wouldn't want to lose.
In the case of the AQ test, I'm guessing that the idea is to tease out an answer from people who feel their position of 'tend to agree very slightly' is closer to 'neutral' than to 'tend to agree'.
-roy
Posted by Roy Badami, Thursday, 21 April 2005 20:03 (link):
I also think it would be interesting to distinguish "neutral" from "don't know", although I'm not sure how you'd use that in the analysis...
There is potentially a big difference between 'don't know/no opinion' where you haven't considered the issue sufficiently or feel you have insufficient information to form an opinion at all, from neutral, where you definitely have an opinion, and it falls in the centre of the spectrum.
In the former case the issue is not likely to be very significant in your decision as to which party/candidate to support. Or at least, how well argued their position is on the issue is likely to be more significant than their actual position.
In the latter case you will presumably seek out a party/candidate which also shares a centrist position on this issue...
-roy
Posted by jdc, Saturday, 16 April 2005 13:40 (link):
Interesting, thanks. My problem with these tests is that I change my mind depending how I interpret the question or how strongly I feel from one day to the next.
I'd be curious to see what would happen if there were a few more questions included on which Liberal Democrat policy is relatively right-wing (minimum wage increases, unemployment reduction programmes, keeping the post office public, trade union rights) and what their voters feel about them.
Equally I worry about some questions - one might not want third world nurses because of being against immigration, or one might not want them because the third world needs them more. One might support GM because companies should be able to do as they wish, or because one has been convinced that it is the only viable way of feeding the world's population in the medium term, etc.
Posted by Peter Clay, Saturday, 16 April 2005 16:10 (link):
I was looking at the axis of UKIP trying to work out what, if any, ideology might underly it, and came to the conclusion that it's about expections of people from different social backgrounds: foreigners, UK underclass, etc., all of whom are assumed to be criminals or spongers by the right and assumed to be decent people by the left.
This is probably not the world's greatest insight, but hey :)
Posted by Chris Lightfoot, Saturday, 16 April 2005 16:16 (link):
Yep. A couple of people have said similar things; the best generic explanation seems to be that it's something to do with the extent to which people fear/loathe outsiders -- who might be foreigners, criminals, poor people, chavs, or whatever. People on the right hand end of the axis think that these people are essentially unlike them and should be dealt with by hanging them, keeping them out of the country, not giving them money, etc. People on the left think that everyone is essentially alike and that these extreme measures aren't necessary. Or something like that. But the dislike of foreigners in particular is culturally specific: the statement `This country should try to become more like France or Germany than the United States of America' comes out as a relatively strong disagree on the `axis of UKIP'. There's probably grist for the `Anglosphere' mill there....
Posted by Chris Lightfoot, Saturday, 16 April 2005 18:58 (link):
Oh, and for what it's worth, I suppose I should link to my results: (-4.7, +2.5). On the `not being a crazed fascist fuck' axis, you will not be surprised to discover that 91% of people are to my right....
Posted by Hadley Wickham, Sunday, 17 April 2005 18:57 (link):
A very nice analysis, but a couple of questions about the principle components analysis:
Did you use any other methods to explore the multidimensional space of the original variables?
Posted by Chris Lightfoot, Sunday, 17 April 2005 19:47 (link):
Two approaches: firstly, bootstrapping from the marginal distributions for each variable, and then computing the distribution of eigenvalues obtained by decomposing the correlation matrix from those samples; and comparing the eigenvalues to those expected from a `broken stick' distribution. The latter approach discards all but the first eigenvector; the former all but the first three or so. But the population doesn't separate on the third axis over any very interesting variables, and it doesn't seem to make much conceptual sense, so I basically ignored it. (Also, as you would expect, bootstrapping from the joint distribution of the variables shows the components of the later eigenvectors become increasingly ill-defined, since after the first couple are subtracted out the residual distribution is basically spherical.)
About 20% and about 7%, respectively.
Yes.
On the results pages from the quiz I only show the first ten because all thirty-two would have taken up too much space on an already-crowded page. There's a complete list of the loadings in graphical form in the slides I link to above, but if you want CSV then here are: questions.csv, which defines the order of the questions (including the voting intention, demographic, and left/right perception ones at the start of the quiz, which aren't included in the PCA), and scale.csv, which has four columns giving the mean, standard deviation, eigenvector #1 component and eigenvector #2 component for each of the 32 values questions (i.e., the last 32 rows of questions.csv).
As for other methods... I had a brief look at multidimensional scaling with distances based on the numbers of questions disagreed on between each pair of respondents. That gives this picture showing a two-lobed structure. But the axes which make up that structure (apart from being rotated wrt those from the PCA) seem to express much the same things, and you can see the same intermixing of supporters of the three major parties. I haven't investigated other approaches in any detail yet, but would be grateful for suggestions. Others have suggested independent components analysis (looking for the most non-gaussian components), but I am not well informed on this and haven't investigated it yet.
Posted by dsquared, Monday, 18 April 2005 08:41 (link):
I would only be inclined to put much weight on the results of bootstrapping from the marginals; I have minimal experience in PCA but seem to remember that the method of comparing eigenvalues to the broken-stick distribution is pooh-poohed in the econometrics literature for having low power and more or less always giving you the first three eigenvalues. But I have a temperamental bias toward bootstrapping and might be misremembering this.
Btw, it's clear to me what the first component is measuring; it's measuring Adorno's F-type personality.
Posted by dsquared, Monday, 18 April 2005 08:52 (link):
Sorry, what the hell am I talking about? It's "broken stick regression" that gets a hard time in econometrics textbooks, which is something more or less completely different. As you were.
Posted by Chris Lightfoot, Monday, 18 April 2005 10:41 (link):
They do look scarily similar on the aspects that relate to crime, punishment, etc. But I'm not sure where the immigration and Atlanticist vs European thing fits in there.
Posted by dsquared, Monday, 18 April 2005 11:25 (link):
Immigration is the elephant in the living room of the f-type; the original work was financed by the ADL or something similar, but the purpose was that it was meant to be a study into the psychological roots of anti-Semitism. My guess is that Atlanticism is a form of "authoritarian submisssion" - the exaggerated and uncritical attitude toward the idealised moral authority of the "in-group". Also that the USA is mentally represented as masculine and Europe as feminine. All very interesting and very deep stuff.
Posted by Sam Evans, Thursday, 21 April 2005 13:51 (link):
I'd have Atlanticism down as a preference for small government rather than the Franco-German statism prevalent in the EU. Small government proponents tend to place a great deal of emphasis on individual responsibility, so tends to correlate with a harsher attitude towards criminals. Immigration may be a kneejerk response to the meme that we are swamped with dirty foreigners living in luxurious council houses whilst scrounging off working folk, and so would correlate more with a dislike of the dependency culture rather than a dislike of foreigners per se.
Posted by Stephen Gray, Thursday, 28 April 2005 14:32 (link):
I'd have Atlanticism down as a preference for small government rather than the Franco-German statism prevalent in the EU.
That's exactly what I took the question to be asking, so I'm pretty sure that plenty of other people did too.
Posted by Hadley Wickham, Monday, 18 April 2005 14:34 (link):
It's probably because it's awhile since I've done any thinking/reading about PCA - but what does statistically significant really mean here? The first two axes only explain 27% of the total variation, so in aggregate the remaining variables must surely in some sense be significant. Is it basically testing for sphericity of the remaining variables? Do you have a reference for the bootstrapping approach?
Other possible methods that come to mind are be finding projections that maximise separation between your classifying variables, "holes" in the data etc. One useful tool to explore this type of highly multidimension data is ggobi (or the in-development site). Unfortunately ggobi is a bit intimidating at first and takes some investment to use well. The ggobi book is a good place to start.
Posted by dsquared, Monday, 18 April 2005 15:02 (link):
so in aggregate the remaining variables must surely in some sense be significant
Not at all; you can extract principal components from white noise - all you'll get is orthogonalised white noise but you can do it. You're right that "statistical significance" is being used in a sense other than the ordinary one but there are procedures you can go through to get an idea of "how big is big" when applied to the principal components of a dataset. The google term is (careful search reveals) "parallel analysis".
Posted by Chris Lightfoot, Monday, 18 April 2005 15:09 (link):
Yeah. I've found this paper by Donald Jackson, Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches, Ecology 74:2204-2214, a useful summary. Here I've looked at approaches #4 and #9, though rather than testing whether successive eigenvalues are distinct (his proposed #9) I used the bootstrapped values as thresholds above which to consider the eigenvalues of the real data significant. In any case in the end I've made a judgment based on what made sense conceptually rather than purely numeric criteria.
(And yes, I am using `significance' loosely here.)
Posted by Hadley Wickham, Tuesday, 19 April 2005 15:35 (link):
That's a really nice summary, although it makes me feel rather defensive of the screeplot (given that that was the way I way taught). I'm still not totally sure about the bootstrap approach for the eigenvector coefficients - don't you need to take into account the arbitrary direction of the eigenvector?
Posted by Chris Lightfoot, Wednesday, 20 April 2005 17:44 (link):
Yes, you do. Basically I match them up by maximising the inner product of the subsample eigenvector with one of the whole-sample ones. You have to account for a reflection as well. This appears to work fairly well, though.
Posted by Chris Lightfoot, Monday, 18 April 2005 15:19 (link):
Yes, but PCA doesn't find any well-defined combinations of them which are themselves significant, so the next few eigenvectors don't supply useful summary information. Reference for the bootstrapping approach -- see the Donald Jackson paper I mention below, though in fact it performs a different test to the one I'm doing (comparing eigenvalues from the synthetic data to those from the real data); this seems a more satisfying test to me.
Posted by Irish Bloke, Monday, 18 April 2005 18:23 (link):
You didn't provide scope for NI voters...
Posted by Chris Lightfoot, Monday, 18 April 2005 18:56 (link):
No, indeed not. YouGov only operate in Great Britain, so (a) we can't compare respondents to the Northern Ireland parties; (b) the model it uses doesn't incorporate any data from Northern Ireland and is unlikely to be very good at describing the opinions of people in Northern Ireland.
Sorry.
Posted by Breandán MacGabhann, Wednesday, 20 April 2005 17:42 (link):
Well, my results came out more or less as I expected: about 95% of Britain is more right wing than me.
http://www.politicalsurvey2005.com/scripts/quiz?R=0;s=KINBLAEGCADDADEDDDABAEEAAEADCCADBEDCBEDBEE
Given this, I was rather surprised to be told that the BNP was the closest match for me on the second axis! Has the BNP got hidden Communist tendencies then?
Posted by Chris Lightfoot, Wednesday, 20 April 2005 18:16 (link):
In a word, yes.
The `closest match' thing looks for the party to whose centre you're closest (strictly, for which the number of people on your left is closest to the number of people on your right). On economics 4% of BNP voters are to your left -- the largest of any of the parties. For comparison, only 0.6% of Lib Dem voters are to your left economically.
Is this reasonable? Well, on the first axis, about crime, immigration and whatnot, you come out as a closet Green and (obviously) quite unlike BNP voters. On the second axis (economics, and stuff) you come out BNP. It turns out that the BNP's economic policies aren't all that far off (say) the Green ones:
A. Yes! You finally understand how the world works. Now close your eyes, make yourself comfortable, and go to sleep. Good night.
The most astonishing bit of this is not that BNP policy on economics and Iraq looks like that of (say) the Green Party, but that its voters somehow know that: the quiz is, of course, comparing your views to those of other voters, not to the parties' policies themselves.
(Two other points: (a) the number of BNP voters polled is pretty small, so there might be quite a bit of sampling error here; (b) the BNP does have an economically far-right -- `free market' -- wing, seen for instance in this plot. Obviously I should also remark that nobody votes for the BNP on economics....)
Posted by Daniel Davies, Wednesday, 20 April 2005 21:43 (link):
This is something that you can have hilarious fun with left-wingers on the Internet, by the way; the underlying truth that Chris's analysis is picking up is that the second word in "National Socialism" is there for a (historic and political) reason. I have at various times flirted with Plaid Cymru, which is both a Nationalist and a Socialist party, and as a result have observed how careful one has to be in having Nationalist and Socialist politics to avoid views that end up carrying you into a theory of Volk rather than class. What is happening here is that the diffusion of political parties in ideology-space is a long memory process, and things that happened in the 1930s and before have echoes that can still be heard today.
Posted by Chris Lightfoot, Thursday, 21 April 2005 23:52 (link):
Hmm. My explanation -- with, I admit, no evidence whatever -- is that the Green/BNP mercantilist position on economics is a sort of ill-informed default. Parties which don't really campaign on economics come up with this stuff in the absence of any actual ideology.
(I'm not, by the way, saying that a theory which relies only on people's natural stupidity is necessarily more plausible than one which relies on `the diffusion of political parties in ideology-space', but I think the possibility should be considered.)
Posted by Hamish Allan, Saturday, 23 April 2005 19:39 (link):
I'd say it would be a bit difficult to argue that the Green Party ended up with anti-consumption and pro-organic economic policies "by default".
War also has a significant environmental impact. I'm quite surprised that the BNP is so anti-war, although the cynic in me says that any party not entrenched in a particular position would be foolish not to get onto that bandwagon.
Not sure where the Greens' welfare policy comes from though. Not that I mind, because at least it gives a left-winger something to vote for.
Posted by Stephen Gray, Thursday, 28 April 2005 14:30 (link):
Not sure where the Greens' welfare policy comes from though. Not that I mind, because at least it gives a left-winger something to vote for.
I think the Greens' welfare policy is just natural development from its other stances - the green movement is generally anti-big business because it sees big business as a major cause of environmental damage. And then there's the "global justice / anti-globalisation movement" which is a big coalition of environmentalists, development organisations, trade unions, and the like, all working together on certain global issues. Basically, the natural constituency of the greens is increasingly committed to social goals and anti-poverty measures, so a left-wing welfare policy is a natural development. It might not flow directly from environmentalism, but it's not hard to see how it appeals to their natural supporters.
Posted by Harald Korneliussen, Thursday, 21 April 2005 17:38 (link):
Sorry if this is a stupid question, but doesn't these axes, and their strengths, depend a lot on what questions were asked? It worries me a little that what we could informally call the fascist axis is all that important ... But if there were fewer questions about immigrants and more about economics, could things be different?
One thing is the methodology of the mathemathics, but I wonder how the questions were decided, and whether, for instance, new questions could be asked to shed more light on the variation in others, so to say.
Posted by Chris Lightfoot, Thursday, 21 April 2005 23:57 (link):
Yes, it does depend on the questions asked. We tried to cover as broad a range of questions as we could, and we did go through two iterations of the poll revising the questions between the two; I think we've done a reasonable job of covering the important areas.
Posted by Jonathan Stanley, Friday, 22 April 2005 01:56 (link):
"Note that Lib Dem voters are significantly to the left of Labour voters on this axis. Partly this is because of opposition to the war against Iraq; partly this is because the Liberal Democratic party no longer resembles is classical Liberal antecedents."
I assume the classical Liberal party you are referring to is the one pre-WWI? Quite how exctly would they differ and how would the Liberal party of yore compare with the Liberal party which split from the LibDems, when the Liberals didi their little merger with the Social Democrats?
Oh and on the scatter graph, any chance of having the leaders and perhaps key movers and shakers of the political parties overlayed?
Posted by Chris Lightfoot, Friday, 22 April 2005 16:14 (link):
Well, the pre-WW1 Liberals were champions of free trade and fought World War I to defend Britain from Prussian Tyranny (or because they were imperialist swine, depending on how you look at it). I'd expect somebody who supported such a programme to take up a positive position on the second axis.
Leaders and key movers and shakers... only if they complete the survey. However, I do have data on people's perceptions of the party leaders, and I'll write something about that later.
Posted by Alan Ford, Friday, 22 April 2005 13:43 (link):
Firstly - Excellent work, Chris, I found this survey - and especially the comparison results - very interesting.
Are you planning on providing the result comparison with data gathered from this survey? I'd be very interested to see how the information changes with a broader sample of people. Of particular interest to me is the plot on the first page.
I find myself (unsurpsingly) strongly in the top-left (mostly libertarian) quadrant of this plot, which I see has a good mixture of voters of all parties in it - it certainly looks more varied than any other quadrant. It'd be great to see this plot updated with a wider sample so it can be seen which parties have greater support in this otherwise fairly sparse quadrant!
Posted by Chris Lightfoot, Friday, 29 April 2005 15:15 (link):
Well, the sample from which the axes are derived is a balanced sample of the whole population, so it should be representative. I will do a plot of where internet respondents come out on the graph too.
Posted by Andrew Marshall, Tuesday, 26 April 2005 21:47 (link):
The questions are too vague. I think working people pay too much tax; i interpret this as tax in general, because VAT is regressive and i disagree with 40% rates of income tax on low middle earners.
Also, private companies are more efficient than public services in many instances, but efficiency is not the motivation for having public services.
Posted by Brendan Hogg, Friday, 29 April 2005 14:50 (link):
Two brief points:
1) The links to the other results pages seem to be missing the "quiz" after "script/" in their URLs, and so come up with a Forbidden error (at least for me).
2) Stephen Pollard provides yet more evidence of his need to be hit with a particularly heavy clue by four here, wherein his coming out very far over on the free market/war in Iraq axis makes him decide the quiz is wrong, because he thinks immigrants are beneficial (he even provides his results link, where one can easily see that he's centrist on the internationalist/rehabilitative axis that measures that).
Posted by Chris Lightfoot, Friday, 29 April 2005 15:06 (link):
(1) -- thanks, I've now fixed that. Out of interest, which browser are you using?
(2) -- god, he is dreadful, isn't he? Still, all publicity is good publicity, as they say.
Posted by Brendan Hogg, Friday, 29 April 2005 15:17 (link):
1) I'm using Firefox under whatever Linux variant is installed on this machine at work.
2) It's such a pity he doesn't allow comments on his blog ...
Posted by Chris Lightfoot, Friday, 29 April 2005 16:36 (link):
Weird. Firefox is the browser I use and the one with which I tested the scripts most. Grrr.
Pollard used to allow comments; he now only allows trackbacks, but I don't think anyone looks at them. I am (childishly) amused to note that his page is still the first Google hit for `ignorant git'.
Posted by sinan, Friday, 29 April 2005 19:16 (link):
Did you try to do a factor-analysis on this data ? it may add valuable information in downscaling your criteria in the sense that the data might reflect that some of the questions are redundant. You can also downscale the objects, it is as simple as the classical PCA (they overlap in some senses).
Posted by Chris Lightfoot, Saturday, 30 April 2005 18:07 (link):
I had a brief look but didn't draw any detailed conclusion. It's on the list of further things to try, though.
Posted by David Boothroyd, Saturday, 30 April 2005 22:29 (link):
Your analysis of politics is way off. Oliver Kamm is certainly not right-wing.
I filled out the survey expressing my preference for UK forces taking international action to remove tyrannous dictators and was told this mean I should be voting BNP! If you read the BNP manifesto you'll find they are isolationist and were against the Iraq war. Removing dictatorships is a left-wing policy, not a right-wing one.
Posted by Chris Lightfoot, Sunday, 1 May 2005 01:39 (link):
I don't think so. What were your results?
He hasn't told me his results but from his writings I'd guess he's about (-1, +4). If you don't understand what that means -- as your comment suggests -- I suggest you re-read what I've written above and the material on the PoliticalSurvey2005 website.
Posted by Peter Clay, Sunday, 1 May 2005 13:40 (link):
Where do you place Galloway and his "respect" lot then? They're what used to be the far left, and are very definitely anti-war. Removing dictatorships by military means seems to be a right-wing policy these days. The Spanish Civil War was a long time ago.
Kamm is difficult to place; I've skimmed his blog and he seems to be pro-Zionist and anti-Communist, which sounds fairly rightwing to me.
Posted by David Boothroyd, Sunday, 1 May 2005 15:49 (link):
Zionism is a left-wing policy.
Posted by David Boothroyd, Sunday, 1 May 2005 15:24 (link):
I've done it a few times; in the latest I scored -3.3, +3.0 (this has been higher). The problem I have is with your axis on "public and private involvement in the economy, international trade, redistributive taxation... and Iraq". You refer to right-wing positions including "The UK was right to go to war in Iraq". This is not a right-wing position - it's a left-wing one: arguing for international involvement to remove foreign tyrannies is fundamentally internationalist. Opposing this position is right-wing isolationism.
The BNP's economic policy is not free-market: they advocate tariffs and economic nationalism. Their foreign policy is to oppose military action to remove tyrannous dictatorships like Baathist Iraq because it's none of Britain's business (echo of Oswald Mosley in 1939 there). Yet when you get a highly positive score on this axis you get told you are like the BNP! This is completely wrong.
Posted by Chris Lightfoot, Sunday, 1 May 2005 21:35 (link):
OK, a question of nomenclature. Ending up on the positive end of the second axis means that you're likely to prefer the railways be in private ownership, think that there is no need to increase taxes on the wealthy, oppose trade tarriffs, and think the private sector more efficient than the public. So far as I can see you object to those positions being called `right-wing'. Well, it's a point of view, I suppose. It turns out that people who see themselves as `left-wing' typically end up with negative values on the second axis, suggesting that trying to claim positive values as `left-wing' is slightly eccentric.
Consider the following:
One of these parties is not like the others. Is it left-wing or right-wing?
Correct. But:
You are being told that you are like the BNP's voters. The analysis knows nothing about the parties other than the views of those who intend to vote for them. Alone among the identified parties in the poll, the BNP has a substantial concentration of pro-free-market voters. (The other parties' voters tail off gradually at large positive values.) Large positive scores on axis 2 plonk you right in the middle of that concentration (see comment, above).
(Admittedly there are not very many BNP voters in the sample, so this classification may be the result of sampling error. But I have no evidence for that particularly, so it would be dishonest to take the easy way out and remove the BNP voters from that comparison.)
Posted by Peter Clay, Sunday, 1 May 2005 22:04 (link):
I think you're missing the way that Chris's "left" and "right" are extracted from the way people's beliefs cluster; so, there is a cluster of people who are in favour of trade controls, redistributive taxation and public involvement in the economy who are against invading Iraq, and this cluster looks like what was once called "left".
Most of the people I've seen opposing the Iraq war and the Occupied Territories look at least slightly left to me. I'm interested to know how you define "left".
(Maybe we need new axes; I propose Strange/Charm, after the quarks)
Posted by Dave, Monday, 22 August 2005 00:05 (link):
Why do people get so wound up about Iraq? I was a British soldier there in 2003. I know the Sheah after so many years of hate with Sadam they are on our side and hate the insurgeants. like usiual we couldn,t finish it without you but we, will fight to the end
God save the Queen
God bless America
Posted by Sean Fear, Wednesday, 14 September 2005 16:45 (link):
That's an interesting test. Only 70% of voters were to the left of me on the first set of questions (Europe/immigration etc), compared to 97% on the second (free markets/Iraq)
WRT some of the comments, I really would be surprised if there was a large cluster of BNP voters out there who are pro-free market. While I doubt if the majority of BNP voters are particularly interested in their party's economic policies, I should think many of them are not that keen on free markets or the Iraq War. On May 5th, the BNP's best votes came in places like East London, South Yorkshire, West Yorkshire, Stoke on Trent, and Lancashire; generally industrial, or post-industrial areas.
I think that that finding is down to sample error. I'm not surprised though that Conservatives and UKIP voters have similar views though.
Posted by Paddy, Thursday, 9 March 2006 20:13 (link):
Interesting test, I turned out moderately to the left on free markets and Iraq and a long way to the left on the hanging and flogging axis. The results showed that I am the typical Guardian a reader.
I'd be interested to know if there's anywhere where I can find out the percentages who said they agreed, disagreed etc with each question. Is there anywhere I can find that out?
Posted by michael daniel, Monday, 19 June 2006 01:20 (link):
I am now convinced that there are, as historically quoted, "lies, damned lies and statistics". Where are the numbers of persons participating? These graphs, which could be manipulated by all sorts of statistical disagreements, completely lose me as a 'floating voter'. Nothing seems to match the qustions I wish to see answered and the graphic curves apparently only show a variation in political viewpoints, not a democratic overview in total.
Post a new comment.
Comments copyright (c) contributors and available under a Creative Commons License. See also the comments policy.