Chart and map from the Abulsme.com 2012 Electoral College Prediction page. Both assume Obama vs Romney with no strong third party candidate. Both show polling as it currently exists. Things will change before election day. On the map red is Romney, blue is Obama, gold states are too close to call. Lines on the chart represent how many more electoral votes a candidate would have than is needed to tie under several different scenarios. Up is good for Obama, Down is good for Romney.
There are two states changing category today. In order by number of electoral votes:
Nevada moves from a “Lean Obama” swing state into “Weak Obama” territory as Obama’s lead in the five poll average goes above 5% to 5.6%. This is still close enough to 5% that a poll or two in Romney’s direction could easily move this back into swing state territory. For the moment though, this means I no longer consider this one of the states Romney could win in his “best case” scenario. So that best case scenario gets slightly worse for Romney, now with Romney winning by just 278 to 260.
Next, Montana moves from “Strong Romney” to “Weak Romney” as Romney’s lead in the five poll average drops below 10% to a 9.6% lead. Now, make no mistake, a 9.6% lead is still very substantial. Montana is a long way from being a swing state. But with a lead under 10% and a trend moving in Obama’s direction, Romney may want to deploy some resources to defend the state even if Obama’s odds of actually flipping the state are very slim (and they are). Now, Montana is only 3 electoral votes… so maybe Romney won’t bother. But with all of Romney’s paths to victory currently being very narrow ones, every electoral vote counts, so Montana will probably get some attention. Because Montana was not and is not a swing state, this does not however change anything in our summary.
| Romney | Obama | |
|---|---|---|
| Romney Best Case | 278 | 260 |
| Current Status | 191 | 347 |
| Obama Best Case | 159 | 379 |
With this new status Romney’s path to victory is holding all of his current Strong and Weak states, then from the swing states winning all of: Florida (29 ev), Pennsylvania (20 ev), North Carolina (15 ev), Virginia (13 ev), Arizona (11 ev), Tennessee (11 ev) and Missouri (10 ev). If he lost ANY of those states, he would lose the election overall. Of the current swing states, he can only afford to lose Iowa (6 ev) or New Hampshire (4 ev)… but not both.
If this was the polling in late October, the chances of Romney pulling off a victory would be pretty slim. (Not zero, a Romney win would still be possible, the odds would just be very strongly against it since it would require a near sweep of the swing states, including many where he is behind.)
It is of course not late October though. Romney has plenty of time to campaign and plenty of time to move critical states toward him.
Edit 2012 May 4 14:05 – Adding in my boilerplate intro paragraph, which I had forgotten.
Have you thought of weighing the polls by age? The oldest Nevada poll is about 6 months old and should really count much in your average.
I have certainly thought about doing some form of age weighting. My reasons for not doing so thus far have basically been A) It makes the calculations more complicated and obtuse, so they would be both harder for me to set up, and harder to understand at a glance, B) As we get closer and closer to the election, the poll frequency increases greatly and this becomes less of a factor, and C) in the few cases I’ve looked at more closely, it ends up not making a huge difference.
For instance, lets look at Nevada.
With the last five polls at the moment you have an Obama lead of 5.58%.
Now, Instead of just doing a straight average of the last five polls, lets do a weighted average of all polls since the 2008 election, weighting them such that the weight decays exponentially with a half life of four weeks… we get an Obama lead of 7.54%. (A two week half life would give 7.81%, a one week half life would give 7.97%.)
Now, this clearly is a little more, reflecting that his more recent polls have been better than the 5.58% average… but it leaves the state in the same category as far as my analysis, so it doesn’t really change anything in terms of my conclusions.
What this WOULD do in general is let the averages react more quickly to changes in sparsely polled states. A lot more weight would be given to the most recent polls compared to the older polls. The “old” situation in the state would be discounted more quickly. In this particular state, nothing changed status, but some states might change. If I had more time, I’d recalculate all 50 states and DC using this method to see how many would be in different categories if I did this. (I don’t think I’m going to have time though… anybody can grab my spreadsheet and try though, it is linked from the wiki linked from this post.)
On the flip side, it would make the estimates much more vulnerable to showing a big move based on one outlier.
For the most part I’ve ended up at the conclusion that in the end the simple five poll average is a reasonable compromise. If your new polls are much different than your old polls, it will take a bit longer to get there while waiting for the old poll to roll off the five, but it will get there soon enough, and once polling gets frequent in the next few months, the simple average will be reacting very quickly anyway in any of the competitive states. (In fact, maybe too much so, if I remember correctly, in some cases back in 2008, there were some states that sometimes got more than five polls in a single day.)
Of course, now that we’re talking about it, I just set up everything I need to calculate this for Nevada, so doing it for all states and seeing what happens is just a matter of doing that for the other states. So it is tempting to do it and see how many states actually do change category…