Over the last few weeks, the Political Compass quiz has been something of a fad among the blogs. It's a descendant of "the world's shortest political quiz", and like it supposedly evaluates you along two independent dimensions, a left/right one and a libertarian/authoritarian one. It's quite useless as a way of eliciting and comparing opinions (since the questions are ambiguous and very leading), but mildly effective as a recruiting tool of the Libertarian Party (since the questions are ambiguous, very leading, and scored so as to make all kinds of people come out as libertarians). The Poltical Compass version is a little better, but not much. Tim Lambert, taking a break from his valuable public service of stalking John Lott, set up a page where bloggers could enter their scores, resulting in a clickable graph of where people come out. The result, interestingly enough, is that almost all of the variation lies along a single axis, running from left-libertarian through right-authoritarian --- though the spread on the libertarian-authoritarian dimension gets bigger as you move rightward. (That is, lefties are almost all libertarian in proportion to their leftness, but there are some rightist libertarians and some rightist authoritarians.) The second dimension of the test is thus actually fairly uniformative.
Chris Lightfoot has taken the initiative to do better than this quite pathetic result. His approach is in some sense obvious to anyone who's heard enough talks on psychological testing. Formulate a fairly large array of questions on diverse subjects; make sure, as far as possible, that they're not ambiguous or prejudical; ask the same question in multiple ways in the same test, to check consistency. Then get lots of people to answer them, and extract the principal components of the answers. (See Lightfoot's notes on the design of his survey.) It turns out that only two of the principal components are significant, and account for most of the variation in the response. The more important of the two corresponds, roughly, to the classical left-right axis. The other significant, but much weaker, component is harder to pin down in conventional terms --- Lightfoot tries "pragmatism/idealism". Three nice features of Lightfoot's method are that (1) the axes are defined endogenously, by the actual variation in people's answers; (2) the axes are actually independent; and (3) by examining the eigenvectors, one can determine, empirically, the "ideal types" of the components --- behold the ideal-typical right-winger and the pragmatist.
Here is Lightfoot's survey, and here is Lambert's new page for plotting bloggers on it; it shows a much healthier two-dimensional spread. (Here is a set of model answers you should all try to emulate.) It would be very interesting to get a large random sample of the population to take the test, and see if the principal components remain relatively stable. My guess is that the leading eigenvector would not change very much, remaining the left-right divide, but I'm not so sure about the second.
Let me go off on a tangent about the basis of the left-right divide. In various forms, this has dominated political self-perception in western countries for several centuries (in England, for instance, since the 1600s), and much of the rest of the world too. Things like these surveys, or more serious attempts at analyzing legislative voting patterns, indicate that it actually does account for a lot of the variance in politics, meaning that people's attitudes and beliefs about lots and lots of different issues are all tied together. My question is whether these positions actually have any logical or psychological coherence. That is, are there reasons why one's views on, say, the death penalty should be related to those on who should build and operate roads? If not reasons exactly, then properties of the human organism which make it hard to hold "divergent" views on such issues? Or is this just an example of self-reinforcing historical lock-in --- that once upon a time, for whatever reason, parties formed which had positions on all these issues, and that since then the social mechanisms of indoctrination, debate and group polarization have tended to preserve the associations? If, to indulge in a thought experiment, we took a population where we randomized their positions across the now-linked issues, so that knowing someone's stance on private roads told you nothing about what they thought of the death penalty, would that population inevitably reconstitute itself along the left-right axis, as people tried to become more coherent, or would it instead polarize along some different constellation of issues? Obviously, no such experiment is going to happen, but I can't think of any other good way of getting at the problem. (Please do not refer me to the political works of George Lakoff. I have my reasons for being exceedingly skeptical of his whole program, but his stuff on politics is especially abysmal.)
Posted at November 15, 2003 23:28 | permanent link