Site Meter

Sunday, October 31, 2010

Washington Post Congressional Ratings

According to The Post's analysis, 19 Democratic-held seats currently lean toward the Republicans, and Democratic strategists all but concede those contests. An additional 47 Democrat-held districts are considered tossups, while 38 other Democrat-held seats, while leaning toward the Democratic candidate, remain in potential jeopardy. Meanwhile, just four Republican-held seats appear truly competitive -- three leaning toward the Democrats and one considered too close to call.

Republicans need to gain 39 seats for a majority. Greg Sargent offers two analyses which don't convince me. I offer a third.

First interpret lean as as solid lead. In practice this fits Cooke's practice, although I have no idea about the W post. Then, first interpret solid lead as certain win, so before getting to tossups the Republicans gain 16. There are 47 Dem held toss ups 46 more than the one Rep held tossup. So if the tossups split 50 50, the Republicans gain
oh my
39 and control the house 218 to 217.

Hmm ok what if lean means 90% chance (I think this is actually close to what happens when Cooke says lean but one should check at fivethirtyeight at a news source which must not be mentioned here).

OK 19 lean R 50 lean D so 31 more lean D. Moving from 100% down to 90% gives Republicans pick up 3.1 more for an expected gain of 42.1 . These numbers are the best numbers for the Dems that I have ever seen. Better, for example, than predicted based only on past elections and GNP growth (Reps gain 45).

Now I'd say the numbers are not really such good news for Dems. I assumed that a huge range of probabilities are called tossups. If so, while splitting toss ups 50%-50% is the only way to interpret the word, the true average probability of a "tossup" could, in theory be an 89% probability of a Republican win.

No comments: