Election Exercise Part III

Election Eve! Here are the final poll numbers for comparison with how the actual count turns out. As in my previous two postings in this series, the estimates are based on the LA Times numbers. Tonight's numbers are in red. Blue numbers are previous numbers.

Method A: States where leads within the margin of error are called ties
Kerry: 153 Bush: 192 Tie: 193
Kerry: 153 Bush: 168 Tie: 217

Method B: All leads are counted as wins (despite margin of error)
Kerry: 269 Bush: 265 Tie: 4
Kerry: 285 Bush: 219 Tie: 34

Method C: Method B with reasonable guesses based on conventional wisdom/history for tied states (Kerry: HI)
Kerry: 273 Bush: 265 Tie: 0
Kerry: 304 Bush: 234 Tie: 0

Method D: 49% rule
Kerry: 280 Bush: 247 Tie: 11
Kerry: 330 Bush: 191 Tie: 17
Kerry: 310 Bush: 182 Tie: 46

Note: This is just for fun. Technically the only method that is statistically significant is Method A. Though not as interesting.

All right - let's see how they do!

BTW - just to be a complete nerd, I also did an analysis that leads me to this prediction of the outcome:

Kerry: 282 Bush: 256

To do this prediction I used a somewhat complicated model where I took the results from the LA Times for states within the margin of error and weighted each state's results by a function of the amount of the lead. The function I used mapped the lead into a cumulative normal distribution based on a margin of error of +-3.5%. Thus even though the states are all within a 95% confidence interval showing no results, my method gives weight based on the theory that the closer you get to the edge of the 95% confidence interval boundaries, the smaller chance that that would occur by chance. Simply put, states with larger leads (even though within the margin of error) get more importance on the final tally based on a somewhat statistically correct hypothesus. Ok - I know most of you don't follow... but lets see how it does.

Below is today's electoral map applied with the 49% rule.