Another aspect of the 2008 returns was found in the Virginia results. Most networks seemed to "hold back" on the Virginia results, and one network (not sure if it was CNN or ABC) used the Virginia results to place Obama over 270 delegates after California closed. (Some Northern Virginia returns came in late, but with other districts in N.Va and the rest of Virginia in, these were a known factor.) The reason was clear -- while the McCain campaign concentrated on Pennsylvania in the last two weeks, it was clear that if Obama could pick up Virginia, he would pick up the rest of the states that were polling in his favour.
In an early "call" of the election, it is helpful to identify "swing states" and other states which are heavily contested:
- "Swing states"
- According to Bruce Haynes of Purple Strategies, the biggest swing states are Ohio, Florida, Virginia, and Colorado. Purple Strategies identifies the "swing states" as:
Arizona, Colorado, Florida, North Carolina, Ohio and Virginia.To this we can add:CNN identifies eight states: Ohio, Florida, Virginia, Nevada, Colorado, New Hampshire and Iowa.
Michigan, Wisconsin and MinnesotaOf these states one or more will be polling solidly for one candidate. If, for example, Michigan continues to poll strongly for Obama, if Romney takes that state, it will be apparent that he will have done exceedingly well elsewhere.
- Other contested states
- These are states whose pre-election polling numbers are close. If one of these states is less favourable regarding a particular candidate then one or more "swing states", that candidate can be expected to win the "swing state". Currently, for Obama, North Carolina is consistently polling less favourably than Virginia and several other "swing states". If the polls continue to reflect this and if North Carolina continues to reflect this difference, then an election day win of "North Carolina would strongly suggest favourable results for Obama in the identified "swing states".
While there are no "other constested states" likely to render results favourable to Romney early in the election, if Romney picks up Pennsylvania or Michigan, this is indicative he will pick up Ohio.
New Hampshire - While New Hampshire may reflect shift national sentiment, the state is isolated from most ad buys and intense campaigning, which affect the "swing states". NH is notable because it enters the "269/269" equation for a tied electoral college vote (which in turn would result in a Romney-Biden administration). (combinations which could lead to a tie; New Hampshire figures in 16 realistic combinations, assuming MO stays with Romney and NM stays with Obama.)
So-far, it seems likely that the key states will be Virginia, Ohio, North Carolina and Florida, with Colorado and Nevada coming in later in the night.
The difficult part of an early election night prediction is the difference between polling results between Eastern swing states and Western swing states. As with the official results, the "call the election" predictive results depend on the math. More later.
Any early-night prediction relies on a combination of pre-election polls and networks calling the results of individual states. Subdividing the pre-election polls there are states that can be reliably predicted to go with one or the other candidate (e.g., Massachusetts for Obama and Utah for Romney), and other states which are close in the polls.
Alternate Versions
1. "Brute Force" Math
This approach takes a combination of large delegate "swing states" and combines these with the states which are aligned with either candidate. The presumption is that if the "swing states" align with the candidate, the other states will likely either shift toward the actual results of the "swing states" or follow the pre-election polls.
This approach is very reliable because most efforts of the campaigns target the "swing states". Therefore, if there are late shifts in voter sentiment, those shifts will be more pronounced in the "swing states".
What to Look For
The swing states are Ohio, Virginia, Florida and Pennsylvania. It appears that at least Pennsylvania will be solidly Obama, but that doesn't shift the math; it just means that the results for PA are pretty-much predetermined. In 2008, Pennsylvania was "called" from the exit polls (when polls closed at 20:00).Looking at the end of September, the following "swing states" are polling in favor of Obama:
- Pennsylvania
- Ohio
- Michigan
- Wisconsin
- Minnesota
- New Mexico
- Arizona.
- Virginia
- North Carolina
- Florida
- Iowa
- Colorado
- Nevada
As an aside, the Pennsylvania results were significant in 2008 because McCain had concentrated late efforts in turning Pennsylvania, counting on Sarah Palin's popular attention in southeast Pennsylvania. (Anyone from SE PA could have told them that did not equate to voter support, but McCain had run out of options.) While Pennsylvania is easy to predict from exit polls, when that state was "called" in 2008, the election was essentially over. The loudest sustained noise I ever heard up until then was at a Richmond, VA post-election party when the PA results came in 30 seconds after the announced close of their polls.In 2012, Virginia may not be as strong of a win for Obama as in 2008, but the results may be more predictable. (It is also likely that Virginia returns will be in within an hour, although Fairfax is still a bit vague about current voting machines.) Therefore, the likely "cliffhanger" could be Virginia or either of the two other closely-contested large-delegate "swing states" -- Ohio and Florida. The reason is Virginia is polling too predictively for Obama and there is not much room for Romney to pick up additional votes in Pennsylvania.The second-loudest noise was the result of the Networks' unwritten rule not to "call" the national election until the West Coast polls closed. (Hawaii takes this one seriously, and closes their polls with California.) Since California, Oregon, Washington and Hawaii were polling solidly Democratic, the only way to do this was to hold back on results from at least one other swing state. Virgina was chosen based on late results from Fairfax County. One of the campaign leaders at the post-election party said "Someone ought to tell them if an election in Virginia is close and you're waiting for Northern Virginia to come in, you already know what the results will be." Well, they already knew what the results would be but made the announcement at the close of the California polls. On at least ABC, Virginia put the results at over the 270 electorial vote threshold for Obama. (This also required the network to "call" California, Oregon, Washington and Hawaii, but those were a given.)
The Flaws
Since the networks will avoid presenting the national election results before 11 PM (23:00) Eastern Time / 8 PM (20:00) Pacific, the results of at least one key "swing state" will be held back. If the missing precincts are identified, it is possible to predict the results of that state and "call" the election earlier based on the network results, but this requires someone with the reqisite skills behind a computer.Still, in 2008, Virgina was "called" by AP, Politico and El Mundo. NPR curiously listed Virginia as "Returns Available", which presumably implied that overseas outlets were already "calling" Virginia. New York Times showed raw results, but that typically does not disclose the results of close states.
Pollster showed a composite of "all nets" and "some nets"., so at 220 electoral votes, one needed only add the 77"electors from CA, OR, WA and HI to get the result.
News Organisations with Early Results for Key States
- Pollster
- a composite of most networks
- Politico
- AP
- Daily Kos (possible)
- El Mundo
- Guardian
- shows results for some states without returns.
- NPR
- shows "Returns Available" suggesting to go elsewhere for an early "spoiler"
- Wikipedia
- New York Times
- A little late, but shows current returns for all.
2. "Fill in the blanks"
This technique is similar to the above "Brute Force" Math methoud,
but relies more on the reliability of the major networks. It is also convenient for
people attending election night pre-hook-up drunks election night
parties.
As mentioned above, the networks will avoid posting the results of one or more key "swing states" prior to the close of polls on the West Coast. This technique takes advantage of this.
How it works
It will be obvious which of the key "swing states" are being held back because the Networks will not be predicting the outcome of the delayed returns. Remember in states having good pre-election polling, the networks are able to use exit polling to "call" the state immediately after the close of polls. This technique should also be available for individual precincts, especially given that other precinct results would already be available. If the state's existing pattern of returns is apparent, it is obvious that the network is able to predict the outcome.Now, take the network's national report and identify the state(s) for which the results are being "held back".
Then go to a source known for announcing early results for key states (list above), and use that source to "fill in" the missing results. If NPR again lists "Returns Available" for some states, that is suggestive that the results for those states are ... available (and not just speculative).
This technique is convenient for an election drunk bash because there is likely to be at least a few people with internet access, who could pull up the "fill in the blanks" websites.
Virginia - In 2008, the Exit Poll News Consortium held back on "calling" Virginia. Virginia was fairly close, with a few precincts, mostly in Fairfax County, coming in late. As someone at the Democratic election night party in Richmond said, "Someone needs to tell the news organizations, if Virginia is close and you are waiting for Northern Virginia, the Democrats have won." Virginia has gone to mostly electronic or scanned-at-polling-station equipment, with similar quick results available for absentee voting, so if Virginia's results are not "called", they are being held back by the News Consortium. (If the results depend on provisional ballots, the decision will be delayed by two or three days.)
3. Shift in Close States
This is based on taking contested states prior to the election, comparing the polls with pre-election polls of large-electorial vote "swing states", and projecting the results to the large-electorial vote "swing states". This works three ways:
- 1. " Combining the results of "swing states" and another close stae is is used to infer the results in other "swing states".
- 2. " The results of one of the selected "close" states whose pre-election polling is significantly less favourable to the candidate is inferred to the remaining swing states.
- 2. " Closely-contested states significantly favouring the other candidate have not been "called".
- This happened in 2008, with North Carolina left "undecided". North Carolina was close but before the election polled less favourably for Obama than other large delegate "swing states". North Carolina was not "called" by most networks before "calling" the national election.
Both scenerios rely on the closely constested state being less favourable to the candidate than the (not-yet-called) "swing state".
This is similar to the mathematical "brute force" calculation except that this presumes that the (not-yet-called) "swing state" went will turn out at least as favourably as the closely-contested state with the "called" results. Start with the presumption that this only works if:
This always works for heavily polled states which decidedly fall with one candidate, but that is not very helpful. Where this matters is in states which are considered "close". If a close state is called shortly after the close of polls, that indicates a verification of the pre-election polls. Conversely, if a state which is close but still consistently polling for one candidate, the lack of a consistent result is suggestive that the polls turned out to be unreliable.
- The pre-election polls reflects the candidate wins, considering the poll's margins of error.
- The the exit poll reflects the results predicted by the pre-election poll (for the selected precincts).
- The differences in the results is consistent with the candidate winning.
- Detailed or accurate polling exists for the state.
How this works
The results for closely-contested states can be used to show a trend. Taking the 2008 results, Pennsylvania was predicted within 5% for Obama, and was the focus of a late campaign push by McCain. Pennsylvania was "called" immediately at the close of polls (I think on ABC, but probably on other networks). While a win in Pennsylvania for Obama would not have been a predictor of the election outcome, the immediate result demonstrated that Obama was getting votes as predicted in the pre-election polls. At that point, it was clear that the election was following the predictions of the polls.
Problems
How this should work in 2012
If a "close" state is "called" based on the exit polls (i.e., "called" shortly after the close of the polls), that indicates confidence in the pre-election polls. If that state was close, then the predictions of the pre-election polls can be deemed reliable.
Problems
While differences in results from polls would be an accurate measure, in most states, there is a delay in receiving the results of some districts. The networks will take this into account when "calling" the election, but that only works for 50%. There are no similar predictions available for other percentages. This means that the actual difference between pre-election poll results and election results cannot accurately be determined until well after the election outcome is known (except in the 2000 election).Stuff left over from 2008
Given the pre-election polls, most states are pretty much a given. For example, if Texas shows up in a poll for Obama, there's really no point in voting for President (at least as far as the election outcome is concerned). It turns out that there are about 10 swing states. Of these, if any of 5 go for Obama, the election is all but over.
Here's the math: If we take the present polling, and remove the following 10 states, that leaves Obama with 250 electorial votes. Since Obama needs 269 electoral votes to win, that means that any combination amounting to 19 votes will win the election for Obama; otherwise it goes to McCain (who needs 270 electorial votes to win).
Looking at the list, Florida and Ohio each have enough electors to bring the total to 250. Three other states, Virginia, North Carolina and MO don't have enough electors, but it's difficult to conceive of these states going for Obama while McCain retains the rest. Therefore, the election can be "called" for Obama from any of the first 5 states in the list.
state electors polls
close
(edt)decisive indicative 2008 election
time "called"
uselectioatlas.org2004 election
time "called"
uselectioatlas.org
VA 13 19:00 * 23:13 20:41 FL 27 19:00 * 23:16 24:27 OH 20 19:30 * 21:23 01:00 NC 15 20:30 * Thursday 20:22 MO 11 20:00 ? 19-Nov 22:16 PA 21 21:00 21:00 22:16 CO 9 21:00 23:35 24:22 MN 10 21:00 21:00 04:38 IN 11 18:00
or 19:00* 02:10: 19:00 NV 4 21:00 21:44 11:34 AM NH 4 20:00 20:00 (called after 11:34 AM)
Indiana is an outlier, but based on 2004 results, we may hear their results before Florida and Virginia. Indiana polls for McCain by about 10 points more than the national average, so if Indiana goes with Obama, there's no reasonable possibility for McCain to win.
The math for McCain is a little more complicated. McCain has a comfortable hold on 164 delegates. Essentially, he has to pick up another 106 but it's easier to look at this in terms of whether Obama can pick up the 19 electors from the above list.
Looking at the numbers, if Obama fails to get the first 3 (Virginia, Florida, Ohio), then North Carolina and IN will also go with McCain. That leaves Colorado, New Mexico, Nevada and New Hampshire. That means Obama would have to get New Mexico and two other states. (He could get Colorado and two other states, but if Obama takes Colorado, he also takes New Mexico.)
This is also the math for McCain, except taken in the negative. If Obama fails to get one of the first 3 (Virginia, Florida, Ohio) then unless Obama takes New Mexico, or Colorado and two other states, McCain wins.
The confusing part is that, if Obama wins any one of the first 5 states
in this list, Biden is elected Vice President. If McCain wins the first 4 states
and particular combinations of Colorado, New Mexico, Nevada and
New Hampshire, Sarah Palin is elected Vice President. We're talking
about a negative here (in the calculation; I leave to the reader to decide which
candidate is 'the negative'), so the final indications if McCain wins the first 4 are
still unclear. Missouri isn't taken into account for McCain because if he takes Ohio
and Virginia, he will also take Missouri.
The times are poll closing times. We can expect exit polls, but in each of the critical states, the numbers will be too close. Regardless, the media will combine exit polls with a few results in key districts to render outcomes shortly after the closing time. During the primaries, Virginia was called for Obama 1/2 hour before the polls closed, so that science is out there.
The way I see it, Virginia will probably be the indicator. Ohio will have its results from Cuyahoga County sometime before 10-December. North Carolina doesn't matter because the results from Virginia will probably be in by the time North Carolina reports, but it is possible that North Carolina will beat Virginia. Florida will probably await the Supreme Court's determination. So if Virginia doesn't go with Obama, it could still be a long night.
The media calling the Virginia primary based on exit polls was an error. (The Northern Virginia polls closed 1/2 hour late due to wx.) Nevertheless, we can expect results substantially before the West Coast polls close at 20:00 (8PM), PST. (Virginia's polls close at 16:00 PST.)
A better example is Maine, in 2004. Their results were "called" when the polls closed at 20:00 (8PM) for Portland and the "At Large" (statewide) districts, but the Northern District (which is also part of the "At Large" district) wasn't "called" until 24:44. Obviously, if the results are close, the "early results" come in late. (In the case of New Mexico, the state was not "called" until absentee ballots were counted a few days after the election.)
West Virginia isn't on the list, but in 2004, Indiana and West Virginia provided the earliest results of any of the states whose outcome is presently uncertain. West Virginia is leaning toward Obama, but I don't think the outcome there is a significant prediction, except for about 5 electorial votes.
McCain
McCain's efforts in Pennsylvania makes a McCain win in this state
commonwealth an indicator of a likely McCain victory. This of course would
require a major shift in the election trends, but similar events had happened
in past elections. The reason I list it as "partially indicative" is
that this would involve a very significant change from the pre-election polling, with
no matching of pre-election polls with election results.
| state | electors | polls close (edt) | partially indicative | 2004 election time "called" uselectioatlas.org |
| PA | 21 | 20:00 | * | 20:49 |