Election Model Update
For the original model writeup, see here.
I made some tweaks to the model, and allowed for a random "drift" in the current numbers, based on the average variance per day (drift variance) in polling data from each individual state. This allows the model to calculate a "forward likelihood" of an outcome assuming the values drift from the current electoral estimate in a stochastic fashion. Drift variance in polling numbers are estimated for each state directly from polling data from the beginning of 2016. For states with no polling data, we average over all known state drift variances and use that value. So we will get two win outcomes: an outcome if the election was held today...and the probability of that same outcome happening in November.
With this in mind the model generates the following outcomes:
If the election were held today, the probability of a Democratic win is
87.3%
and the probability of an R win is
12.7%
with an average of 308 electoral votes, and a median of 310 for the Democrats. The most likely outcome for Dems is 322 EVs
####################################################
The probability of a D win in November is:
73.6%
and for the Rs:
26.4% (of course)
with an average of 289 electoral votes, and a median of 291 for the Democrats. The most likely outcome for Dems is 296 EVs.