Heretics' almanac: Bush or Kerry with days to spare

Slate's political editors have taken the trouble of collating all the state opinion polls to guess who would win the election, as of today.

They've sorted states into seven buckets, like this
- Out of play, going to Bush: Alaska, Idaho, Utah, Montana, Wyoming, North Dakota, South Dakota, Nebraska, Kansas, Oklahoma, Texas, Louisiana, Mississipi, Alabama, Georgia, South Carolina, North Carolina, Tennessee, Kentucky, and Indiana
- Out of play, going to Kerry: California, Illinois, District of Columbia, Maryland, Delaware, Connecticut, Rhode Island, New York, Vermont
- Bush likely: Nevada, Arizona, Colorado, Missouri, Arkansas, Virginia, and West Virginia
- Kerry likely: Oregon, Washington, Minnesota, Michigan, Pennsylvania, New Jersey, and Maine
- Bush, but iffy: New Mexico, Florida, Iowa
- Kerry, but iffy: Wisconsin, Ohio, New Hampshire, Hawaii

The result, according to their analysis is that Kerry now has 149 solid electoral votes, 85 likely, and 38 iffy, for a total of 272. That would win him the presidency compared to Bush's 168 solid, 59 likely, 39 iffy for a total of 266.

But, Kerry has fewer locked up than Bush does, so it makes sense that both are trying to make inroads in the "iffy states," i.e., Wisconsin (10 electoral votes), Florida (27), New Mexico (5), Iowa (7), Ohio (20), New Hampshire (4), and Hawaii (4). (Though I haven't heard of either of them visiting Hawaii or Alaska).

Slate makes a call based on their judgment on the so-called battleground states.

I tried a different approach, assigning a probability to each state. Relying on Slate's interpretation of each state, I assigned a probability that it would go to one candidate or another. I gave each of them 100% for the states that are "out of play," 80% for the "likely" states, and 60% for the "iffy" ones in their favor. The opposing candidate got the remaining probability.

I then multiplied the electoral votes by the probability for each state. This is virtually guaranteed not to predict the electoral vote outcome, but may give a risk-adjusted prediction for the overall outcome. (I say "may" because the uncertainty attendant to say Florida has a much greater effect than the uncertainty for Iowa. In other words, we're not comparing apples to apples."

With that approach, Kerry gets 261.6 electoral votes, and Bush gets 276.4, and the election goes to Bush.

The electoral vote page also calls it for Kerry, though with the caveat that two states - Michigan and Iowa, amounting to 24 electoral votes - are too close to call. Applying a similar weighting technique to this page, things look even worse for Kerry with only 227.54 electoral votes.

Electoral Vote also links to an interesting article in the Washington Post (registration required) that explains why polls may be completely misleading.

Another geek point: if electoral college votes were reapportioned to states according to their population, Kerry would get 272 electoral votes with my analysis on the Slate data and win the election. It would also mean that California would gain 10 electoral votes, and several states (Alaska, Delaware, DC, North Dakota, South Dakota, Vermont, and Wyoming) would only get to elect 1.

Nordic perspectives

General rolls