Brad DeLong posted a link to a Political Coordinates Test the other day. I see these tests a few times a year – figure out where you lie along Left-Right and Liberal-Conservative axes. This one is similar to ones I’ve seen previously but that they have a little twist. While they have a Left-Right axis instead making the second axis a Liberal-Conservative one they use a Liberal-Communitarian continuum instead. Conservative and Communitarian are aren’t the same thing – at least not in the context of US politics. Here’s a sample plot with some well-known political figures for reference.
Perhaps more interesting than where it places you on their 2-D political spectrum is the test itself. Many of their questions leave room for interpretation. Take for example, “The market is generally better at allocating resources than the government.” I read that and think, “The market is generally short-sighted, is frequently non-rational, and lacks the capacity for large-scale coordination of effort on complex tasks. The market would completely screw up national defense, infrastructure, funding for basic research, etc. To the extent that it’s insinuated itself it’s screwing those things up more than they already were. I can trust the market for things which don’t require long-term planning or coordinated allocation of resources or to allocate resources for things which aren’t particularly important (e.g., what choices of potato chips are available to me at the supermarket) but is the market generally better at allocating resources? No. There are some areas where I believe it’s better and others where I believe it would fail miserably. On things like national defense, infrastructure, funding for basic research it would almost certainly allocate resources differently than the government but I don’t believe it would do a better job.” So I answered “Disagree” to the question. That stated, I can imagine someone reading the same question and thinking, “Communists eat @#$%!” – a sentiment with which I concur – and answering “Strongly Agree.” I bet there’s a correlation between how one interprets questions and where one falls in the Political Coordinates square but the ambiguity in the questions presumably introduces some variation in the bottom line scores.
That begs the question, how do they distill the responses to their 36 questions down to a 2-D score? I suspect commenter Jorgensen on DeLong’s blog has it right:
I think that this is the sort of test you create by:
(1) coming up with questions,
(2) asking a few hundred people from as broad a selection as you can find,
(3) run a singular value decomposition on the matrix of responses
(4) look at what the first two singular vectors are
(5) try to understand what the first two singular vectors represent in common political language.
Beyond SVD, Isomap and Locally-Linear Embedding also came to mind as a possibilities for projecting a non-planar structure down to a 2-D plane. That’s getting a bit into the weeds however.
Anyhow, I took the test three times over four days. Median scores: 58% Left and 22% Liberal. Curiously, that’s about the same as Obama scores. So why don’t I care more for him? That’s left as an exercise for the reader.