Dear Kossacks,
Here is an update of a calculation I first posted on this site two weeks ago. The permanent link is here and includes all the details.
It's a snapshot of many recent state polls taken all at once. The basic idea is to use the statistical variation among polls to calculate the likelihood that Kerry or Bush will win a state. This is similar to the margin-of-error concept, except that it takes into account not only sampling variation, but also variations among pollsters' methods.
I then calculate the likelihood of
every possible outcome among 17 battleground states and Tennessee (which, statistically, has come into contention). From these nearly 300,000 possible combinations I estimate the overall probability of a Kerry win if the election were held today and reflected those polls.
Results as of Wednesday, August 4, 2004
Counting the last three polls for each state, the current probability of a Kerry win is 90 percent. 270 electoral votes (EV) are needed to win. The median (50th percentile) outcome is Kerry 291 EV, Bush 247 EV. The 95 percent confidence band for Kerry is 264-318 EV. Most battleground states have had at least one poll since the Democratic convention.
The rank order of battleground states is currently (boldface indicates states statistically in play at the 20-80% level):
D <- MI/OR/PA/WA/WI/NM/NH/MN/IA (95-100%) - ME - TN - WV - FL - MO - NV - AZ - OH/AR (0-5%) -> R
These calculations would be affected by an overall poll bias. In a close race, even a small bias can have a large effect. Bias could happen if sentiments have changed since the time of the polls, or if state polling methods sample a population that differs from actual voting patterns. For instance, increased motivation by Democrats would cause a bias (one that we like) by elevating turnout on Kerry's side.
The effects of bias on the calculation are as follows.
One point favoring Bush: Kerry win 58%, 50th percentile 275 EVs (95% confidence 255-307 EV).
One point favoring Kerry: Kerry win 99.2%, 50th percentile 308 EVs (95% confidence 275-333 EV).
This leads to an interesting interpretation that matches intuition: as of today, the race stands at a point where a one-point differential in turnout can make a very large impact.
Sam