State-wise electoral analysis

This week Newsweek (Aug. 9 print edition) introduced a new weekly feature called "Swing State Watch." Given the nature of the electoral college, the state-by-state analysis is considerably more relevant than national opinion polls.

Newsweek refrains from predictions, but I have put the Newsweek data into my own newly created model which weights electoral votes by how solidly they are leaning toward each candidate. The model cranked out a prelimnary result: Bush squeaks by Kerry 272 to 266 electoral votes.

But first, back to the Newsweek data. Since I cannot find a link, I'll summarize the findings here:

• Total Bush: 25 states, 211 electoral votes
• Total Kerry: 15 states+DC, 207 electoral votes
• Toss-up: 10 states, 120 electoral votes

The data are further broken down as follows:

Solid Bush (15 states, 110 electoral votes)

• Alabama, 9 electoral votes
• Alaska, 3
• Idaho, 4
• Indiana, 11
• Kansas, 6
• Mississippi, 6
• Montana, 3
• Nebraska, 5
• NDak, 3
• Okla, 7
• SCar, 8
• SDak, 3
• Texas, 34
• Utah, 5
• Wyoming, 3

Likely Bush (7 states, 72 electoral votes)

• Ariz, 10
• Ark, 6
• GA, 15
• Kent, 8
• LA, 9
• Tenn, 11
• Va, 13

Leaning Bush (3 states, 29 electoral votes)

• Colo, 9
• NCar, 15
• WVa, 5

Solid Kerry (7 states+DC, 92 electoral votes)

• Conn, 7
• DC, 3
• Hawaii, 4
• Ill, 21
• MD, 10
• Mass, 12
• NY, 31
• RI, 4

Likely Kerry (3 states, 73 electoral votes)

• Calif, 55
• NJ, 15
• Verm, 3

Leaning Kerry (5 states, 42 electoral votes)

• Dela, 3
• Maine, 4
• Mich, 17
• Oreg, 7
• Wash, 11

Toss-up (10 states, 120 electoral votes)

• Fla, 27 (49% Bush, 45% Kerry)
• Iowa, 7 (46, 45)
• Minnesota, 10 (45, 48)
• Missouri, 11 (48, 48)
• Nevada, 5 (46, 43)
• NHamp, 4 (45, 47)
• NMex, 5 (42, 49)
• Ohio, 20 (44, 49)
• Penn, 21 (43, 48)
• Wisc, 10 (42, 48)

The Newsweek piece is a bit lacking in explanation of methodology. For example, it is unclear if the Kerry/Bush percentage breakdown for toss-up states is based on likely voters or registered voters. It does state that all polls were conducted between July 6 and July 22. Also, the terms "solid," "likely" and "leaning" are not defined.

Back to my analysis:

I've done another stage of analysis and created a preliminary model which predicts Bush beating Kerry, 272 electoral votes to 266.

Assumptions in my model:

• The electoral votes are weighted by how solidly they lean toward each candidate. Each candidate wins
• 90% of his "solid" electoral votes and 10% of the other guy's;
• 80% of his "likely" electoral votes and 20% of the other guy's;
• 70% of his "leaning" electoral votes and 30% of the other guy's.
• For the toss-up states:
• I normalized the percentages to so that the total of each state is 100%.
• I then assigned each candidate a weighted average of the toss-up electoral votes, weighting each vote by the percent likelihood that the candidate wins that vote.
• All data are based on the Swing State Watch from the Aug. 8 print edition of Newsweek.
• This methodology is somewhat flawed:
• The model gives each candidate roughly half of each toss-up state's votes. In reality each toss-up state will go fully to one or the other candidate. The model's implicit assumption therefore is that the total electoral votes of the toss-up states split roughly in half (or more exactly, 59 to Bush and 61 to Kerry).
• The model is based on data from Newsweek which are not fully explained. I do not know if the toss-up state figures are based on registered or likely voters, and I do not know the definitions of "solid," "likely" and "leaning."

My model breaks down electoral votes among Newsweek's classifications as follows:

• "Solid" states: Bush 108, Kerry 94.
• "Likely" states: Bush 72, Kerry 73.
• "Leaning" states: Bush 33, Kerry 38.
• Toss-up states: Bush 59, Kerry 61.
• Total: Bush 272, Kerry 266.

And thus ends my analysis for the day.

(By the way, for folks wanting to do back-of-the envelope state-by-state electoral analysis, I recommend OpinionJournal's electoral college calculator.)