It's the BCS, stupid.

I just read this excellent column in the Wall Street Journal. Not only do I feel cool for knowing how the math works, but certain others who are more politically involved than I will probably get a kick out of me acknowledging that their interests bear fruit once in awhile. The article is about using past history (strength of schedule, if you will) and assumptions about the importance of the economy to predict the outcome of this November's presidential election. It doesn't look good for the Democrats; unless, of course, the model is based on erroneous assumptions. You can check it out for yourself here. Unfortunately, the only explanation accompanying it is an exhortation to buy this guy's book, but don't worry, I know plenty about regression and believe me, there is nothing to see here.

Update: there is an explanation, and I wish you luck in understanding it. It includes the data from which the formula was derived, but there is still no reasoning given as to why he chose the variables he did. The equation has a decent track record, however (save for 1992), so I guess we'll just have to wait and see if the predictions will be borne out.

My take: I have my doubts, given how much politics have changed in the last hundred years--I tend to think that voters today act differently and according to much more complex sets of variables than those laid out here. Given that they failed in 1992, and 2000 probably didn't count, misses in more recent years don't bode well for these assumptions. Thoughts, anyone?