Friday, January 11, 2008


USA Presidential Election: self-test fun!


This morning I indulged in a binge of USA Presidential Elections quizes. I took all six listed at the University of Michigan's Document Center; I had taken two several weeks ago: the ABC News Match-O-Matic (not on the list but the USAToday one seems almost identical) and the WQAD quiz. They vary in length from eleven to twenty-eight questions. Most quizes are composed of questions of the "choose one position from the options we provide" type, but there were also "how strongly do you oppose or support this position" ones (glassbooth, votehelp), and a few instances on some tests where multiple answers could be selected (age ranges, or profession experiences, for instance). All, I think, allow one to indicate how important one considers a question or issue and adjust the weighting accordingly.

A rationalisation

The triggering event: someone asked me how strongly I supported or agreed with one of the candidates, so I spent some time and energy finding out. But honestly, I enjoy the narcissism and vanity of these tests! To check that they all give the same result or, if not, I have some idea why not, gives added relish to taking a bunch of tests instead of just one.

I will not be voting in the California primary on February 5th, so it is somewhat of a waste of time to be considering all the candidates, over and over, instead of waiting to see who the parties nominate. I could pretend to have "evaluated" the various matching tools, but that wasn't what I was doing -- I didn't answer all of them in a way sure to be consistent, I didn't take them several times each with various value profiles simulating different voters in the multi-dimensional opinion space.

What did I find out?

  • The winner ranked first four times, second once, third once. No other candidate came near: top three on every one of the six tests (no other did better than four tests), first in the majority of cases (two others were in the top rank once each). But this was not the candidate about whom I was asked.
  • Place (second best) on the combined results, I reckon, was the candidate that took one first, one second, and two third places. This was not the candidate about whom I was asked, either.
  • "Show" (third best) is a bit arbitrary. One candidate ranked second three times and third once, so was in the top three, four times. This beats the one with three third-places but no firsts or seconds. However another ranked first once and third once: a "first", but only in the top three on two of the six tests -- apparently not a very robust match -- and this is the candidate about whom I was asked! Depending on how much more weight I give to a "first" versus a "second" or a "third", this candidate might get third place on the combined results, but might only deserve fourth or fifth!


The candidate about whom I was asked matches some of my opinions, more than any candidate from the opposing party. However, that candidate is not as close a match as two other candidates, and only about as close as yet two more, so my enthousiasm is not overwhelming.


As if I had been evaluating some aspects of these tools:

  • VA Joe : good test, but my over-all number two was not among the top four in this result.
  • : good test, but my over-all number two was not among the top four in this result.
  • USA : slightly odd result, but I didn't use the weighting adjustment possibilities. Candidates one and two (in my synthesis) were there in my top three, as was one of the contenders for third, but the order was different. Macromedia Flash user interface is "fun", but the final result is only a ranking, without coefficients of matching or similarity.
  • : short, simple; results consistent with the overall combined results.
  • : the result that surprised me. I don't really understand why it happened.
  • : I don't like the formulations of several of the questions (too vague, too general): "more money for education"? It should depend on how it would be used, shouldn't it? And I only answered 20 of 28 questions.

Identificateurs Technorati : , , ,

Powered by ScribeFire.

StumbleUpon Toolbar Stumble It!
You write very well.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?