10:00 AM – That’s all from us
No live blog of the results today . If you want to follow the results, http://electoral-vote.com/ will be running a live blog, but rest assured there are plenty of other outlets for election analysis.
If you want to follow the vote count by county. Check out our guides to the vote patterns for each swing state here.
Bye for now.
PS : Electionate is running a live blog of the results. Highly recommend Nate Cohn.
9:55AM – Betfair & Intrade
If you can’t wait for the networks to project winners then have a look at these two sites. Election wonks that are happy to risk their cash will be looking to wager their hard earned in these two markets, but they won’t be doing it blindfolded. They will be looking at booth by booth returns and other indicators so these market may plunge towards 100% odds for one candidate before the networks call the election. Here are the links:
9:40 AM – How the US networks call elections
In Australia we are used to booth-matching as the tool to eliminate bias from partial election results and give us a prediction of the final outcome by applying the booth matched swing uniformly to the previous final election results.
In the US they do things a little differently. The US Television networks do not have the advantage of a single data feed that provides results by polling booth across the entire country for historical and regulatory reasons. Remember we are watching 51 different elections held on the same day. Therefore the US Television networks jointly commission a single organisation Edison Media Research (EMR) to provide election prediction analysis services.
These services include the conduct and provision of exit polls, but also the same organisation uses a technique called “sample precincts” to provide further analysis and potentially, the early call of elections. The sample precincts technique is the (more complex) American version of booth matching.
In each state EMR sends out workers to sample of polling booths (precincts) to obtain the final vote count and quickly send the results back to the EMR national centre.
The vote results in the sample of booths are compared with previous elections at those same booths. The previous presidential election is not just used, several previous Presidential, Governor and Senator elections are matched with the incoming vote figures for the sample precincts to find which previous election is the best fit with the results.
The exit polls for a given state are conducted in a subset of the sample booths so that an estimate of the vote for each of sub-sample booths can be used just prior and at poll closing to give the sample precincts analysis a head start. As the actual vote results are phoned in for these sub samples of booths they replace the exit poll estimate of results in that booth and shrink the margin of error of the sample precinct estimate.
Furthermore by having workers at the booths as the actual vote results are released and immediately transmit those results to the EMR national centre this provide a faster reporting of the booth results than the official state government channel, which requires the booth workers to set the results to their county, which in turn send the results to their local secretary of state office, who may or may not publish the results by booth.
Therefore the speed at which elections are called in the US is not solely due to statistical techniques it is also due to a logistical advantage in reporting the results.
That being said sample precincts are just that, a sample, which has its own sampling error. If an estimate of a race has a lead less than 5 points, the networks wait until fuller returns are posted by county to call an a election. A model of the county vote for given state is used to call elections on this basis.
Again, if the estimate result using the county model indicates a lead of 2% or less, the networks are now unwilling to call a result on the night. In 2004 and 2008 states with a final margin of 2% or less were not called on election night, with the exception of Indiana in 2008.
9:10 AM – Exit Polls
In about 2 hours the US networks will start releasing exit polls when the voting ends in various states. The important aspect of these polls is that they are not the same as phone polls or regular face to face polls in that they are not a simple random sample of voters.
Exit polls which have a polling day component ( all states bar Oregon and Washington which have all-mail ballots) interview voters that leave a sample of polling booths in a given state. This means the sampling is two fold, a first sample of polling booths are chosen and then a sample of voters leaving those booths are taken. The technical term for this is cluster sampling.
Because of the two fold random sampling process here, the margin of error for exit polls is up to twice that of a regular polls.
A rule of thumb for calling races of exit polls is a margin of 8% or more between the candidates. Of course US networks are bit more cautious than that as they average the exit polls with a pre-election estimate of regular opinion polls for a given state, when they publish exit polls on their websites at poll closing time.
8:55 AM: What if all undecideds break to Romney?
If the polls are overstating Obama’s performance then it is possible that a hidden Romney vote could be in the undecided column, or to put it another way, if undecided voted voters break totally towards Romney how would the EV projection change.
If we assume a 1% non-major party vote, then Obama would need to have 49.5% support to withstand the full allocation of the remaining undecided vote to Romney.
Below is a table of our seven day poll average with the vote shares of the major candidates.
The state where Obama is leading in our poll average but not breaking 49.5% are Virginia, Iowa, Colorado and New Hampshire. If these all fell to Romney the EV tally would be a tight 271-267 Obama win with the president carrying Ohio by 0.6 points.