Here's some of what's being said about this campaign -- from a public opinion perspective -- as we enter its final few days: Gallup leans too Republican. Ohio and Iowa polls have too many Democrats. Obama can't be winning if Independents are going with Romney by double digits. The national numbers and the state polls are telling two different stories. All of the above are probably true. But instead of creating confusion, our sense is that it's leading to some clarity. With five days to go before the election, it is now pretty clear that -- barring a destabilizing external event -- Mitt Romney will likely win the national popular vote for president. The only remaining question is whether the president's Midwest firewall will hold and give him an Electoral College victory.
More than anything else, we strongly believe in three predictive indicators, all of which suggest that the president will lose the popular vote. First is his current vote share, which stands at approximately 47-47.5 percent. Incumbents rarely improve on their final poll numbers by more than one or two points, putting Obama at approximately 48-49 percent on Election Day.
Second, Romney has more than a 10-point lead among Independents nationally and in most key swing states. Our average of national polls fielded over the last two weeks shows Romney with a 14-point lead among Independents. For context, that's a full six points better than Obama's eight-point advantage over McCain in 2008. The Ohio poll average shows Romney with an eight-point lead among Independents. The only outlier in the below chart is Kerry's 19-point win among Independents in 2004 (something that may be attributable to "quirky" exits that year).
It is highly unlikely that the president can win this election if he is losing Independents by double digits. And there aren't any indicators suggesting that Democrat turnout will be large enough to compensate for that.
Which leads to the third point: voter enthusiasm and energy is with the GOP. A recent Pew poll had Republicans with a 14-point lead over Democrats in terms of vote likelihood.
Put these three elements together -- Obama's current vote share, Romney's lead among Independent voters and overall GOP enthusiasm -- and it strongly suggests that Romney will win the popular vote. But if you want some more empirical data... here you go.
Using a regression equation, we were able to produce a model trend and project Romney's vote share out to election day. As of today, the model trend says that Romney is +1.9 percent. Tomorrow he would be +2.2 percent. Friday is +2.5 percent. The intercept is almost exactly +3.5; in other words, the below model suggests that Romney will win the popular vote by 3.5 points. Now let's assume this trend flattens a bit because of Super Storm Sandy. Our current estimate (which we will update next Tuesday morning) suggests that Romney will capture 51 percent of the popular vote to Obama's 48.5 percent. The trend line-based on 26 national polls conducted over the last 30 days --is both unmistakable and virtually unassailable.
Here is our up-to-the-minute take on the current environment:
All of the above is telling us one thing: this race is exceedingly close. We said a few weeks ago that this will come down to Ohio and we stand by that. The polling in Ohio shows the president up two to four points (and some have the president leading by even more than that). The question is whether Democrats are being oversampled; our gut instinct is that they are. If they are not, then Obama wins the state and wins the Electoral College quite handily. Both camps have a strong "ground game" in the state, but our inclination is to look at voter enthusiasm and Republicans have the edge there. Wisconsin, Iowa, Colorado and NH are also too close to call at this point. Nevada is leaning Obama and Virginia and Florida toward Romney.
On Tuesday morning we will release our final popular and electoral vote projections.
Thanks again to Pete Ventimiglia, Kelsey Cohen, Chris Blunt and Allison Quigley for their insights and contributions to the Election Monitor. Follow us on Twitter: @Steve_Lombardo.
Please note that the author was an advisor to the Romney for President campaign in 2008, but is not affiliated with any campaign in 2012.
Follow Steve Lombardo on Twitter: www.twitter.com/Steve_Lombardo
 |
 |
|
| Obama | Romney | |
|---|---|---|
| Electoral Votes (270 to win) |
332 | 206 |
| Obama | Romney | |
|---|---|---|
| Total | 65,899,660 | 60,932,152 |
| Percent | 51.1% | 47.2% |
| Democrats* | Republicans | |
|---|---|---|
| Current Senate | 53 | 47 |
| Seats gained or lost | +2 | -2 |
| New Total | 55 | 45 |
| Democrats | Republicans | |
|---|---|---|
| Seats won | 201 | 234 |
"Trend Compass".
TC is a new concept in viewing statistics & trends in an animated way
by displaying in one chart 5 axis (X, Y, Time, Bubble size & Bubble
color) instead of just the traditional X and Y axis. Discover trends
hidden in spreadsheets. It could be used in analysis, research,
presentation etc. In different business sectors, to name a few we
have Deutsche Bank, NBC Universal, RIM, Vanguard Institutional
Investor, Ipsos, Princeton University as our clients.
Link on Drilling feature (Parent/Child) - Just double-click on any bubble:
http://www.epicsyst.com/test/v2/drilling/
NBC presentation on TED using Trend Compass exported Videos on CNN
Money/Fortune:
http://tech.fortune.cnn.com/2010/12/21/from-laughing-to-judging-in-fifty-years-the-evolution-of-televised-emotion/
Link on our new Geographical Trend Compass (Earthquake in Japan - Mag
vs Depth vs Time):
http://www.epicsyst.com/test/v2/japan-earthquake
Link on Ads Monitoring on TV Satellite Channels.
http://www.epicsyst.com/test/v2/tv-monitor/
Link on UK Master Card vs Visa performance :
http://www.epicsyst.com/test/v2/mastercard_vs_visa/
Regards.
Amr Salah
Trend Compass Team
Epic Systems
www.epicsyst.com
Is that a cubic polynomial fit over four months of data? Way to destroy all meaningful trends in the polling, Steve.
Excel can't magically fit the correct curve to a set of points, it can only determine how well the data fits the model selected. By the look of it, Lombardo selected a "polynomial" trendline. When you do this in Excel, it lets you choose the "order" of the polynomial. A second order polynomial has one "hump" - a U-shaped line (or an upside-down U) - the curve trends upward on either side. A third-order polynomial has both a maximum and a minimum - an N shape. The line trends downward on one end, upward on the other. This graph starts low, hits a max around -90, a minimum around -50, and then shoots upward. Probably a third-order polynomial. But a fourth-order curve change direction once again. Do that and you'd probably get a peak around -20, and then the curve would turn downward again.
So what's the actual trend here? To answer that, you need to get into the guts of the analysis and figure out which curve fits best (something Excel doesn't let you do). Without more information, this curve is meaningless. Since Lombardo provided no information about how good the fit was, rather than being "unmistakable and virtually unassailable" it's "inconclusive and virtually uninterpretable".
I'm partisan
I found a 'model' that fits my wishes
summary: your conclusion has no basis in reality
http://www.gallup.com/poll/156446/2008-obama-voters-mccain-voters-switching-sides.aspx
If you instead give Romney 12% of Obama '08 voters, you still only get Romney 65.4 million to Obama's 64.6 million. Any way you slice it, the popular vote is a toss-up at this point.
You could then make assumptions about enthusiasm, but remember that some '08 voters have passed on (and they were older and more white than the electorate at large), while there are new voters in the process (and they are younger and more racially/ethnically diverse than the electorate at large). Enthusiasm may favor Romney (although it's anti-Obama enthusiasm, and anti-incumbent enthusiasm is never as strong -- see Bush in '04), but demographic shifts favor Obama.
And then there's get-out-the-vote efforts. Few doubt that Obama has the advantage in this department, the only question is will it be enough of an edge to hold onto his small but consistent leads in most of the swing states. I would say that's a good bet at this point, but the wild card is how the storm's aftermath will throw a wrench into the equation.
Why doubt half their data but hold up the other half as sacred truth? If you think all the polling is off, then they could just as easily be off on the indy vote.
Even just over the last few years, response rates to polls have plummeted and cell phone-only use has sky-rocketed. These trends should favor Republicans, even if you employ demographic weights (ie, young people with landlines skew more GOP). You can see this by comparing polls that handle cell phones differently.
Also, partisan affiliation has historically proven to be very fluid, not a surprise when you consider that most states don't register voters by party, or do so oddly (ie, what primary you last voted in, even though this year the only competition was on the GOP side). This includes most swing states, most notably Ohio.
Since pollsters (other than Rasmussen) are not starting with a preset partisan breakdown, and are MEASURING partisan self-identification ALONG WITH presidential preference and likelihood to vote, this "Dem oversampling" is probably a reflection of the president's supporters simply identifying with his party to a greater degree than the challenger's supporters identify with the opposition party -- which is the situation we saw in 2004.
Dems learned that the hard way since back then it was THEM doubting the partisan breakdown of the polls.
Congratulations, Steve, on compressing the least amount of insight into the largest number of words.
F&F
But seriously, I can't wait for the end of the elections. Not because I want to compare your results against the traditional pundits, never mind the right wing ones. But more on I'm excited how your model would fare against the other models (538, rcp, etc.), based on the test you proposed.
Please please include unskewed polls as well, just for fun.
the popular vote advantage mitt held weeks ago (a point or two) has been steadily decreasing to fractions now (either way depending on analysis)...rassmussen even confirms this
yet the graph above shows his lead increasing and at an ACCELERATED rate to boot!
Second, when you average the national polls over the last couple of weeks, it is Obama that is trending up, not Romney. The upward trend for Romney that you get for a larger time window is driven strictly by the post-Debate 1 spike, and that spike has long ceased and reversed.
Third, state polls provide more accurate data and therefore better indicate the state of the race, anyway. It is more likely that the national polls will move in the direction of the state polls rather than the other way around.
Obama will top 50%.
Also, author Lombardo says NH, WI, IA and CO are toss-ups and Florida and Virginia favor Romney. On what planet? Has he looked at the polls lately? Obama has clear leads in the four "toss-up" states, he also leads in VA, and Florida is a complete toss-up.
One can only conclude that this is wishful thinking masked as statistical science. If all those polls prove to statistically biased toward Team Blue, then Lombardo could be right. Maybe But that's very unlikely. The consistency among the polls, and there are a LOT of them, is striking.