RELIABILITY.

The Texas primary has been called for Hillary Clinton; meanwhile, the Texas caucus is still being counted with Obama in the lead. The differential is nearly the same in each contest, about 4%. Same electorate, same candidates, eight-point swing. Does this make sense? Is this ridiculous Texas system the perfect illustration of why our primary model is screwed?

In social science, we have an idea called reliability. The idea is that you take different measures of the same thing to be sure you're measuring it right. For instance, I might measure somebody's propensity to use blogs for surveillance reasons -- that is, to observe and follow news stories -- by asking whether they use blogs to "learn how politicians stand on issues" and to "help me make up my mind about things." Using a statistical test, I can find out how much these questions seem to be tapping the same underlying idea (in this case they went with four other items, and had a Cronbach's alpha of 0.844). If an election is meant to provide an estimated measure of the "will of the voters," something is unreliable in Texas -- they're coming up different. I don't mean to say that the outcomes are necessarily statistically different -- 52C-48O and 48C-52O are fairly close -- but that these two processes, ostensibly designed to measure the same thing, appear to have different outcomes. That's kind of crazy! Texas has inadvertently given us a field experiment comparing the caucus and primary processes, and given us different results. I find that very interesting as a social scientists, but as a small-d democrat, I find it both baffling and disturbing.

UPDATE: A handful more precincts have reported in, bumping Obama's caucus lead to 12 points with 39% of precincts counted. And that will be it, because the rules in Texas make precinct reports voluntary. But rest assured, superdelegates are the real problem in this system.

Posted by Aaron S. Veenstra ::: 2008:03:05:13:50